Welcome to Day 26.

Everyone is posting. Posts scroll past, and they’re gone by lunch.

The thing people actually keep, save and come back to is a resource. A checklist. A simple guide. A little tool that solves one real problem. Today you make one, and then you do the part that matters: you give it away.

Want to see what you’re aiming for? Here’s the AI Fluency scorecard I built, up and running.

Watch: the AI Fluency scorecard in action (60 seconds)

And it’s live, so you can try it yourself: switched-on-scorecard.vercel.app. That’s the bar we’re aiming for today.

A word before we start

This is the generosity move, and generosity is how you get remembered. Visibility is not shouting louder. It is being the one who handed someone something genuinely useful. People remember who helped.

And it sits on the right side of your line. AI builds the thing, the layout, the structure. The judgement about what is actually useful, and any words in your voice, those stay yours.

One rule for today, because it’s the one we keep breaking: a resource that never leaves your folder isn’t a resource, it’s a draft. Today’s job isn’t finished when you’ve built it. It’s finished when someone else has it.

Step by step

Step 1.
Pick the one problem. Not your whole expertise. One thing your audience, the real woman from Day 18, keeps getting stuck on. Small and specific wins. “How to brief a designer.” “The five questions before you hire.” “Am I actually using AI well?”

Step 2.
Open Cowork in your desktop app, pointed at your Switched-On folder. Opus for this one. It reads your audience and pillars, so the thing you make fits the people you actually serve.

Step 3.
Pick the format. Two routes, choose by how brave you feel:

• The reliable one, for everyone: a one-page guide or checklist. It comes out as a clean PDF or a Canva design, ready to send, print or post. No faff, no hosting, nothing to set up.

• The wow one, if you fancy it: a simple interactive tool, a scorecard or quiz that gives the user a result. More impressive, a little more to build.

Ask Claude to help you choose:

Read my audience and pillars. I want to build a free resource that solves this one problem for them: [the problem]. Suggest the best format, a one-page guide, a checklist, or a simple interactive scorecard, and tell me why it fits my audience. Keep it to something I can finish and share today.

Step 4.
Build it.

Build it for me. If it’s a guide or checklist, make it a clean one-page PDF on my brand. If it’s a scorecard, make it a single web page I can open and share. Leave any line that should be in my voice marked CHANGE THIS so I write those myself. Don’t render anything heavy until I’ve seen it and said yes.

The AI Fluency scorecard up top is the one I made, the standard to aim for: genuinely useful, not a secret advert.

One practical bit, because it trips people up. Made a one-page guide or checklist? You’re done, just attach it or post it, no hosting needed. Made an interactive tool and want it on a link to share? The no-code way is a drag-and-drop host like Tiiny.host or Netlify Drop: upload your file, get a web address, send it. That’s all my scorecard link is, a file put online.

Step 5.
Make it yours. Write the CHANGE THIS lines. Then read it as her: would she actually use this? Would she thank you for it? If not, tell Claude what’s missing and go again.

Step 6.
Give it away. The whole point of the day.

• Post it: “I made a thing that solves [the problem]. Comment and I’ll send it over.”

• Or offer it as a free download in your newsletter.

• Or send it straight to one person who needs it today.

Add to MEMORY.md: Day 26, built a free resource that solves [the problem] and gave it away. AI built it, the useful judgement and the voice are mine.

A safety note

Useful, not salesy. A resource that is secretly an advert gets binned in seconds. Give the real thing, the genuinely helpful version, and trust the goodwill to come back.

And nothing with private or client data in it. What you make today is built to be shared.

What you’ll notice

The first reply. “Ooh, I need this.” “How did you make that?” That’s the moment a post never gives you. You didn’t just take up space in the feed. You were useful in it.

Why this works

A post is borrowed attention. A useful thing is earned trust. One scrolls past, the other gets saved and passed on with your name attached. AI has collapsed the cost of making the useful kind, the kind that used to be expensive. So make those, and give them away.

Key takeaway

Stop making things that vanish by lunchtime. Today you made something people keep, and you put it in their hands. That is what being chosen is built from.

Where this is heading

Tomorrow we make the whole thing run without you, and get properly precise about what is actually an agent and what isn’t.

Tomorrow on Switched On

Day 27: Set it running, and walk away.

Put your recipes on a schedule so the work happens while you sleep. And finally get straight on the difference between a recipe, a scheduled task and a real agent.

Nishma x

P.S. How did today land? Three buttons, thirty seconds. Tap here

I read every response. The data shapes the rest of the challenge.

Catching up? Days 0 to 25 are on the Switched On Substack archive HERE.

Got a friend or a colleague who this might be useful for? Send her to nishma.co/30daychallenge

#SwitchedOn #SwitchedOnAIChallenge

​Unsubscribe | Update your profile | 167-169 Great Portland Street, London, W1W 5PF

Welcome to Day 25 of the challenge.

For three weeks you’ve been filling a folder. Who you are. How you sound. Who you serve. How the money works. What you’re known for. The decisions you’ve made. Some days it felt like admin. It wasn’t.

Today you find out why. You point Claude at the whole folder and have a conversation you can’t have with anyone else, because nobody else has read all of it at once.

A word before we start

A chatbot starts cold every time. Your folder means Claude starts knowing you. The gap between “give me advice” and “given everything you know about me, what would you say” is the gap between a search engine and a counsel.

One honest framing. This is reflection, not prophecy. A sharp mirror, not a fortune teller. Use it to see yourself more clearly, then you decide.

Step by step

Step 1.
Open Cowork in your desktop app, pointed at your Switched-On folder. Opus for this one.

Step 2.
Let it read everything, then hold up the mirror. Paste this to Claude:

Read every file in this folder. You now have my about-me, my voice, my audience, my business, my pillars and my decisions. Before I ask you anything, tell me the three things you notice about me that I might not see myself. Be honest, not flattering.

Step 3.
Sit with what comes back. It will name a pattern you have been living inside and not quite seeing.

Step 4.
Now ask the big ones. A few worth your time:

Given everything you know about me, what should I focus on for the next ninety days, and what should I drop?

Where am I inconsistent? Where does what I say I value not match what my files show I actually do?

What am I sitting on? The thing I keep circling but never commit to.

What is the one throughline of my whole year that I haven’t named yet?

Step 5.
Keep what lands.

Save the sharpest things from this conversation as a dated note in my folder. These are the ones to act on.

A safety note

This folder is the most personal thing you’ve built in this challenge. Keep it local, keep it yours.

And remember the mirror only reflects what you fed it. If a file is thin or out of date, the answer will be too. When something feels off, that is usually the file talking, not you. Update it and ask again.

What you’ll notice

It says something true that you’d half-buried. Not because it is clever, but because it read all of you at once, and you almost never do. Most of us hold our own story in fragments. Today you see it whole.

Why this works

Every file you built this month was an investment. Today it pays out. This is the whole reason context matters: the same question, asked of a tool that knows you, gives an answer worth having. You did not build a tidy folder. You built a thinking partner.

Key takeaway

A chatbot answers questions. A second brain answers YOUR questions, because it knows you. You spent three weeks building one. Use it for the hard thinking, not just the busy work.

Where this is heading

Tomorrow we make something with all of this. Not a post. A real thing your audience can use.

Tomorrow on Switched On

Day 26: Build something your audience can use.

A resource, a tool, something genuinely useful with your name on it. The kind of thing people keep.

Nishma x

P.S. How did today land? Three buttons, thirty seconds. Tap here

I read every response. The data shapes the rest of the challenge.

Catching up? Days 0 to 24 are on the Switched On Substack archive HERE.

Got a friend or a colleague who this might be useful for? Send her to nishma.co/30daychallenge

#SwitchedOn #SwitchedOnAIChallenge

​Unsubscribe | Update your profile | 167-169 Great Portland Street, London, W1W 5PF

Welcome to Day 24 of the challenge.

This week you’ve watched two agents do real jobs. The morning brief that gathers your day. The industry review that watches your world. Both mine, built to a pattern.

Today you build one that’s yours. And to show you how, I’m going to build mine in front of you: the one I reach for most, the one that turns a post I’ve written into a finished carousel. Then you build yours, the same five parts, for whatever eats your time.

A word before we start

Every agent, however clever it sounds, is the same five parts. The trigger, when it runs. The job, what it does. The context, what it reads to do it well. The output, where the result lands. The guardrail, what it must never do.

Mine takes a post I’ve already written and lays it out as a carousel, on my brand, ready to upload. I write the words. It does the design. That line, my words and its build, is the whole point. The AI doesn’t write in my voice. It frees me from the two hours in Canva.

Pick yours by the groan test. The task you sigh before starting. Repetitive enough to hand over, important enough to want done well. That’s your candidate.

Step by step

Step 1.
Pick your agent. Mine’s the carousel machine. If your groan is a different one, three that suit most people:

• The carousel machine. You write a post, it lays it out as an on-brand carousel, slide by slide, ready to upload. Your words, its design.

• The inbound drafter. People ask you the same kinds of questions by email or DM. The agent drafts a reply in your voice, filed for you to check and send. It never sends.

• The weekly money check. Every Friday it looks at the model you built on Day 19 and flags anything tight, anything overdue, anything worth chasing.

Pick the one that makes you groan the loudest. I’ll build the carousel machine here, and the shape works for any of them.

Step 2.
Open Cowork in your desktop app, pointed at your Switched-On folder. Opus for this one.

Step 3.
Build it and test it on something real. Here’s the exact prompt I used. Swap the job for yours and the shape holds:

I want to build an agent that turns a post I’ve written into a finished, on-brand carousel. Let’s build it the five-part way and test it on a real post now.

First read my brand files in this folder, my colours, fonts and wordmark, so it comes out on brand.

The spec:
Trigger: I run it whenever I’ve written a post I want as a carousel.
Job: take my post exactly as written, find the natural slide breaks (slide 1 the hook, one idea per slide, the payoff near the end, a simple call to action last), and design each slide.
Context: my brand files and the post. My words, my brand.
Output: build it as one HTML file with each slide at 1080 x 1350, then save every slide as its own numbered image (01, 02, 03) in a new folder so they upload in order.
Guardrail: do not change or rewrite my words, lay them out only. If a line is too long for a slide, tell me and let me cut it myself. Never invent a stat or a quote.

Here’s the post: [paste your post]. Build it, then show me the slides.

Step 4.
Run it by hand first. Same rule as the others. Watch it work in front of you before you trust it. When it flags “this line’s too long for slide 3, want to cut it?” instead of quietly rewriting you, that’s the guardrail doing its job. Tweak until the slides are the ones you’d actually post.

Step 5.
Save it as a one-tap job.

That’s the one. Save it as a reusable agent in a file called carousel-machine.md in this folder, with the full five-part spec, so next time I can say “run the carousel machine on this post” and paste a new one.

One line from now on: “run the carousel machine on this post”.

Step 6.
Lock it in.

Add to MEMORY.md: Day 24, built my own agent, the carousel machine. The pattern is five parts: trigger, job, context, output, guardrail. I write the words, it does the build. I can make these myself now.

A safety note

Your own agent will be closer to your real work, so the leash matters more, not less.

It builds, it never posts. Nothing goes public, gets sent, spent or deleted without you. For the carousel machine that means it makes the slides and stops. You’re the one who hits upload. Those words go in the guardrail and stay there. And know how to switch it off before you switch it on.

What you’ll notice

The quiet click of “I can build these”. Up to now you ran agents I designed. Today you designed one. That’s the line between using AI and directing it, and you just crossed it. And in the carousel’s case, the two hours in Canva, gone, with your words still your own.

Why this works

Once you’ve seen the five parts twice, the third is only filling them in for your own life. The pattern is the skill. Any single agent is just one use of it. And there’s no ceiling on it: the same five parts that build a one-tap carousel can build a whole app. You keep the pattern forever, long after these particular jobs change.

Key takeaway

Three agents now, and the third is yours. You do not need me to build the next one. You have the recipe.

Where this is heading

Tomorrow, the quiet payoff of building this folder all month. You get to talk to it.

Tomorrow on Switched On

Day 25: Talk to your second brain.

Three weeks of files in one folder. Point Claude at all of it and ask the questions only someone who knows you could answer.

Nishma x

P.S. How did today land? Three buttons, thirty seconds. Tap here

I read every response. The data shapes the rest of the challenge.

Catching up? Days 0 to 23 are on the Switched On Substack archive HERE.

Got a friend or a colleague who this might be useful for? Send her to nishma.co/30daychallenge

#SwitchedOn #SwitchedOnAIChallenge

​Unsubscribe | Update your profile | 167-169 Great Portland Street, London, W1W 5PF

Welcome to Day 23 of the challenge.

Staying current has quietly become a second job.

The scroll that starts as research and ends as forty minutes gone. The newsletters you subscribe to out of guilt and never read. The low hum of FOMO that you’re missing the thing everyone else has seen. You’re not behind. You’re just drowning in inputs, like everyone, and calling it keeping up.

Watching your industry is a job. Yesterday you built one agent. Today you build a second, and hand the watching over.

This one looks outward. Your morning brief watches your day. This one watches your world.

A word before we start

One firm line, because it’s the line that makes this challenge different.

This agent gathers, sifts and scores. It tells you what happened and where the gap is. It does not write your posts, your opinions or your take. That’s the whole point. The internet is already full of AI saying the obvious thing in the same flat voice. Your edge is your view, and your view is yours to write.

And it scores what it finds against the pillars you built yesterday. So the review comes back curated to you, ranked by what you actually care about, not a generic news roundup. It surfaces and sorts. You speak.

Step by step

Step 1.
Tell it what your world actually is. Vague in, vague out. Be specific:

1. The handful of names you genuinely want to keep an eye on. Competitors, peers, the two or three voices who actually shape your space.

2. The topics that matter to your work, in plain words.

3. The sources you trust, if you have them. Particular newsletters, publications, sites. If your newsletters land in Gmail, it can scan those too.

Jot these down. You’ll feed them in. The pillars file from yesterday does the rest, it’s what the agent scores everything against.

Step 2.
Open Cowork in your desktop app, pointed at your Switched-On folder. Opus for this one.

Step 3.
Build the review and run it by hand first. Same rule as Thursday, see it work before you trust it. Paste this to Claude:

I want a weekly review of my world, scored against what I actually care about.
First read my pillars.md, so you know my themes.
Here’s my world: the people and competitors I watch are [names], the topics that matter are [topics], the sources I trust are [sources]. If my newsletters land in Gmail, scan the last week of those too.

Then give me:
1. A one-line summary: how many sources you scanned, how many you skipped as noise.
2. The ideas worth my time. For each: which pillar it fits, a fit score out of 5, a hook in my voice, one line on why it’s mine, the source, and how urgent it is (this week, this month, or evergreen).
3. Three themes building across the week.
4. The white space: what nobody in my world is saying that I could own.

Gather and score only. Don’t write the posts or the opinions for me.

Read it. If it’s too broad, narrow the names and topics. If it scored something too high, tell it why. Shape it into the briefing you’d actually want to open on a Sunday.

Step 4.
Save it as a one-tap job. Works for everyone, today. Paste this to Claude:

Save that exact review as a reusable prompt in a file called industry-review.md in my Switched-On folder, with clear instructions.
Then each week I can open a fresh chat, point at it and say “run my industry review”.

One line a week from now on: “run my industry review”.

Step 5.
Optional upgrade, if your app can schedule it. Set it to run once a week and land where you’ll read it, Sunday evening or Monday morning. If scheduling isn’t simple in your setup, the one-tap version is plenty. Same as Thursday: keep the gather-and-score-only guardrail in the instructions, and know how to switch it off before you switch it on.

Step 6.
Lock it in. Paste this to Claude:

Add to MEMORY.md: Day 23, built my second agent, a weekly industry review. It scans my sources, scores ideas against my pillars, finds the white space, and never writes my opinions for me. One tap, or scheduled.

A safety note

Treat what it gathers as a tip-off, not gospel. AI summarising the web can get a detail wrong, miss context or repeat something that isn’t quite true. Before you act on or repeat anything it surfaces, especially about a named person or competitor, check the original source. Gather fast, verify before you speak.

And keep the guardrail: gather and score only. The day an agent starts publishing opinions in your name is the day your voice stops being yours.

What you’ll notice

The scroll loosens its grip. When you know a tidy, scored briefing is coming once a week, the compulsive checking eases, because the fear of missing it is handled. You get the signal without the hours of noise.

You’ll also notice the white space line is the gold. Anyone can tell you what happened. The genuinely useful prompt is what nobody’s saying yet, because that gap is exactly where your next idea, your next post, your next bit of standing out, lives.

Why this works

Keeping current is mostly gathering, filtering and ranking, slow, repetitive, easy to do badly and exhausting to do well. That’s precisely the work to hand a machine. What it can’t do is have the take. So we split it cleanly: the agent does the watching and the sorting, you do the thinking. Each does the half it’s actually good at.

Key takeaway

Your second agent, looking outward. It watches your world, scores what it finds against your pillars, and hands you the ideas that matter and the gap nobody’s filling. It gathers and sorts. You decide what to say.

Two agents now. One on your day, one on your world. Both on a short leash. Both working while you don’t.

Where this is heading

This week we’ve made things visible and handed jobs to agents. Next week we go further. We take the watching you set up today and turn it into the engine that feeds your ideas, the one that hands you angles only you could write. Plus the bit a few of you have been asking about: what happens after Day 30.

Tomorrow on Switched On

Day 24: The digest that clears the noise.

A third small agent, and the cleanest desk you’ve had in ages.

Nishma x

P.S. How did today land? Three buttons, thirty seconds. Tap here

I read every response. The data shapes the rest of the challenge.

Catching up? Days 0 to 22 are on the Switched On Substack archive HERE.

Got a friend or a colleague who this might be useful for? Send her to nishma.co/30daychallenge

#SwitchedOn #SwitchedOnAIChallenge

​Unsubscribe | Update your profile | 167-169 Great Portland Street, London, W1W 5PF

Welcome to Day 22 of the challenge.

Week 4 opens, and the theme is visibility.

Your bio is a job title and three buzzwords.

We’re all taught to describe ourselves in the safe, sanded-down language that fits a LinkedIn headline and says nothing. Founder. Strategist. Passionate about growth. It could be anyone. It says nothing about what you actually stand for or why anyone should care.

You’ve spent a week building a folder that knows the real answer. Today we turn it into something people can see.

Two jobs. First, find the three things you’re genuinely known for, in your own words, from the evidence. Then build a real page that holds them.

A word before we start

A quick reminder of the line from Day 16, because today it matters.

When the words are your voice, you write them and Claude sharpens. When the work is the thing, Claude can build it and you approve. Today is both, in the right order. Claude does the heavy lifting, finding the patterns, building the page, making it look good. But the handful of lines that have to sound like you, the sentence at the top that’s the truest thing about you, those you write or rework yourself. Claude assembles the page. You own the voice on it.

Don’t publish a page of AI words about yourself. Build the page, then make the voice yours.

Step by step

Step 1.
Find your pillars. Not what you’d like to be known for. What you already are. Open Cowork in your desktop app, pointed at your Switched-On folder. Opus for this one. Then paste this to Claude:

Read about-me, audience and business. If I have writing samples in the folder, read those too.
Tell me the three or four things I’m actually known for, the beliefs I keep returning to and the ground I clearly own, based on the evidence, not on what I wish were true.
For each one, give it a plain name and a single sentence in my voice.
Push back if I try to claim something the evidence doesn’t support.

Step 2.
Argue with it. The first pass is a starting point, not the answer. Be honest. Paste this to Claude:

That third one is the brand I’d like, not the one I’ve earned. Replace it with the theme my evidence keeps circling.
And the wording on the first one is too smooth, here’s how I’d actually say it: [your words].

Push until the three lines feel like things you’d defend out loud, in your words.

Step 3.
Save them. Paste this to Claude:

Save these as pillars.md at the top of my Switched-On folder, and reference it in CLAUDE.md: See pillars.md for the three things I’m actually known for.

Step 4.
Now build the page. The version that works for everyone, today. Paste this to Claude:

Build me a single-page personal site as one HTML file I can open in a browser.
Clean, modern, uncluttered, the kind of page a sharp independent advisor would have.
A strong headline at the top, a short line on who I help and how, my three pillars each with its sentence, and a simple way to get in touch.
Leave the headline and the opening line as placeholders marked CHANGE THIS, I’ll write those myself.
Save it as my-page.html.

Notice the placeholders. The page is built for you. The voice at the top is written by you. That’s the line, kept.

Step 5.
Make the voice yours. Open the page. Find the CHANGE THIS lines. Write the headline yourself, the truest, sharpest version of what you’re about. If you want help, ask Claude to edit your attempt, not to write it from scratch. Paste this to Claude:

Here’s my go at the headline: [your words].
Sharpen it, keep it mine, don’t make it sound like a brochure.
Give me three options that are all clearly me.

Pick one. Or write a fourth that’s better. Drop it into the page.

Step 6.
Optional, the designed version. If you’d rather have something for social than a web page, and your Canva connector is set up, ask Claude to build the same three pillars as a short, clean carousel in Canva instead. Same rule: the words that are your voice, you write. Claude lays it out.

Step 7.
Lock it in. Paste this to Claude:

Add to MEMORY.md: Day 22, found my three pillars from the evidence (pillars.md) and built a one-page site. The voice lines are mine, Claude built the page around them.

A safety note

What goes on a public page is public. Before you put this anywhere people can see it, read it as a stranger would. No private numbers, no client names you haven’t cleared, nothing you’d not want a journalist or a competitor to read.

And keep the line: don’t publish AI-written voice copy under your own name. Build with AI, speak for yourself.

Where to use it

This is the one link you point everything at. The bio line on LinkedIn, your email signature, the thing you send before a call instead of a CV. Most people’s “about me” is scattered across platforms they don’t control. This is the one you own, and the one you send.

To make it live, drag the file onto a free host like Netlify, or get someone to point your domain at it. Not ready for that? A screenshot still works as a “here’s me” you can send today.

What you’ll notice

Naming your three things is harder than it sounds and more freeing than you’d expect. The relief of it. For the first time, what you’re about fits on one page and sounds like you, not like everyone.

You’ll also feel the jump from private to public. All week you’ve built files only you can see. Today something exists that other people could look at and immediately get you. That’s visibility, and it started with knowing exactly what you stand for.

Why this works

Most personal branding fails because it starts with aspiration, the brand you’d like on a good day, which wobbles because you don’t quite believe it. Pillars drawn from the evidence hold, because they’re already true. You’re not deciding what to be known for. You’re noticing what you already are and saying it out loud.

And building the page proves the point of this whole challenge. AI did the part that’s fiddly and slow, the layout, the structure, the look. You did the part only you can do, the truth at the top. Art and science. That’s the partnership.

Key takeaway

Three pillars, found in your own words, not chosen from a wish list. One real page that holds them. The voice on it, yours. The build, Claude’s.

You went from a folder only you can see to something the world can.

Where this is heading

Tomorrow we build the agent that keeps you current without the endless scroll. The one that watches your industry so you don’t have to, and tells you where the gap is.

Tomorrow on Switched On

Day 23: The agent that watches your industry so you don’t have to.

Your second agent. Set it once. Get the week’s moves, and the white space, every week.

Nishma x

P.S. How did today land? Three buttons, thirty seconds. Tap here

I read every response. The data shapes the rest of the challenge.

Catching up? Days 0 to 21 are on the Switched On Substack archive HERE.

Got a friend or a colleague who this might be useful for? Send her to nishma.co/30daychallenge

#SwitchedOn #SwitchedOnAIChallenge

​Unsubscribe | Update your profile | 167-169 Great Portland Street, London, W1W 5PF

Welcome to Day 21 of the challenge.

Friday, the end of Week 3.

This week you built the brain. Not files for the sake of files, the real thing: who you are, how you sound, who you actually serve, how the money really works. About-me, voice, audience, business, all sitting at the top of your folder, read on every session. Along the way Claude tidied a folder you’d been avoiding, built you a money model, and took over your morning catch-up.

Today the brain earns its keep on the hardest job of all. Helping you decide.

On Day 16 I named three ways to use AI. Editor, when your voice is the product. Author, when the work is the product. And a third we saved for later. The challenger. The one who argues against you so you decide better. Today’s the day.

A word before we start

Nobody warns you about this part of being senior. The more respected you get, the less anyone pushes back.

Your inner circle is fond of you. Your team wants to please you. The people around you are, mostly, too polite to tell you the brave thing is the wrong thing, or that the safe thing is fear in a smart coat. So you carry big decisions alone, turning them over in the shower, weighing the same options for weeks.

Claude isn’t polite. It has no stake in pleasing you and no fear of the room. Hand it the decision and the whole brain you’ve built, tell it to argue the other side properly, and you get the thing you’ve been missing. A worthy opponent. You still decide. You just decide with the strongest version of the counter-argument in front of you, instead of the weak one your worry invented.

Step by step

Step 1.
Name the decision you’ve been carrying. The one your closest people have heard you weigh more than once. Should I drop the service that pays well but drains me. Do I hire. Do I raise prices. Do I say no to the big client who’s a nightmare. Do I finally start the thing.

Get it down to one clear question.

Step 2.
Open Cowork in your desktop app, pointed at your Switched-On folder. Opus for this one. The brain files you built this week are the whole point today, this is why they live in this folder.

Step 3.
Set the challenge. Paste this to Claude:

Read my brain files first: about-me, voice, audience and business.
Then I want you to be my challenger, not my cheerleader.
Here’s a decision I’ve been weighing: [the decision, and the option I’m leaning towards].
First, argue hard for the opposite of what I’m leaning towards. Make the strongest possible case, the one a sharp, fond, honest colleague would make.
Then tell me the blind spot you can see in how I’m framing this, given what you know about me and the business.
Don’t soften it. I’d rather hear it now.

Step 4.
Sit with it. Don’t fire back straight away.

Let it land. The good challenge feels slightly uncomfortable, that’s the sign it’s working. If the counter-argument is weak, push: “that’s the polite version, give me the one that would actually worry me.”

Step 5.
Now bring it home. Paste this to Claude:

Now drop the opposing hat.
Given everything, my situation, my blind spot, what actually matters to me, walk through how you’d weigh this.
Lay out the two or three things the decision really turns on.
Then ask me the one question I need to answer honestly to make the call.

Answer the question. That’s usually where the decision reveals itself.

Step 6.
Capture it, so future-you remembers the reasoning, not just the outcome. Paste this to Claude:

Save this as a note in a decisions folder: the decision, the strongest counter-argument, my blind spot, and what I decided and why. Date it.

A safety note

Challenger mode is for sharpening your thinking, not outsourcing it. The decision is yours. Claude doesn’t know your gut, your circumstances or your appetite for risk the way you do. Use the argument to test your reasoning, not to replace it.

And keep the big, life-shaped decisions, health, money you can’t afford to lose, anything legal, with the real professionals too. Claude is a brilliant sparring partner. It is not your doctor, your lawyer or your accountant.

What you’ll notice

The relief of a fair fight. For weeks the decision felt heavy because you were arguing with yourself, and you kept letting yourself win, or letting fear win. Put the strongest counter-case on the table and the fog clears. Either you hold your position with more conviction, or you change it for a good reason. Both are progress. Circling is the only bad outcome.

Why this works

We don’t make worse decisions because we lack information. We make them because nobody shows us the strong version of the argument we don’t want to hear. We defeat a weak straw man in our heads and call it deciding.

Claude builds the steel version instead. And because it has read your brain files, the challenge lands on you, your actual situation and patterns, not a generic version of the problem. That’s the difference between advice off the internet and a colleague who knows you.

Key takeaway

The third mode, and the most powerful one for women doing senior work. AI as challenger. It argues the other side, names your blind spot, then helps you weigh it. You decide.

Three weeks in, the brain you built is now the thing in the room that’s honest enough to disagree with you.

A quiet word about what comes after

We’re two thirds of the way. Next week we make all of this visible and put it to work in the world.

A lot of you have been replying to ask what happens when the thirty days end, whether there’s a way to keep going together. There is, and I’ll share it properly towards the end. For now, just know I’m building something for the women who want to carry on past Day 30. More soon.

Where this is heading

Tomorrow Week 4 opens. Visibility. We stop building the brain and start building things people can actually see. First up, the one page that says what you’re about, and the three things you’re truly known for, found in your own words.

Tomorrow on Switched On

Day 22: Build the one page that says what you’re about.

Your pillars, surfaced from the evidence. Then a real, visible thing that holds them.

Nishma x

P.S. How did today land? Three buttons, thirty seconds. Tap here

I read every response. The data shapes the rest of the challenge.

Catching up? Days 0 to 20 are on the Switched On Substack archive HERE.

Got a friend or a colleague who this might be useful for? Send her to nishma.co/30daychallenge

#SwitchedOn #SwitchedOnAIChallenge

​Unsubscribe | Update your profile | 167-169 Great Portland Street, London, W1W 5PF

Welcome to Day 20 of the challenge.

Every morning you rebuild the day from scratch.

Open the calendar. Scan the inbox for anything on fire. Try to remember the three things that actually matter today underneath the noise. You do it before you’ve properly woken up, and you do it again tomorrow, and the day after.

That reassembly is a job. And jobs can be handed over.

Today you build your first agent. Don’t let the word scare you. An agent isn’t code, and it isn’t clever. It’s just a job you set up once and get back done, again and again, without asking each time.

A word before we start

Every agent, however fancy it sounds, is five simple parts. Once you see them, the mystery goes.

The trigger: when it runs. The job: what it does. The context: what it reads to do it well. The output: where the result lands. And the guardrail: what it must never do.

That last one is the most important, and it’s the same line you drew on Day 16. Your morning brief drafts and summarises. It never sends, never replies, never deletes. It puts the day in front of you. You decide what to do with it. AI does the gathering. You stay the boss.

We build it in that order, and we test it by hand before we ever leave it alone.

Step by step

Step 1.
Make sure it can see your day. The brief is only as good as what it can read. If you connected Gmail on Day 7, that’s one source. A connected calendar is the other. If you haven’t connected your calendar, you can still do today, you’ll just paste your day in by hand.

Step 2.
Open Cowork in your desktop app, pointed at your Switched-On folder. Opus is fine here.

Step 3.
Build the brief and run it by hand first. This is the bit people skip and shouldn’t. See it work in front of you before you trust it. Paste this to Claude:

Build me a morning brief.
Read my calendar for today and tell me what’s on, back to back or with gaps.
Scan my inbox and flag only the few things that genuinely need me today, ignore the noise.
Then remind me of my top three priorities (you’ll find them in my notes, or just ask me).
End with one short nudge: the single most important thing to protect time for.
Keep it short enough to read standing up.
Draft only. Do not reply to anyone or send anything.

(If your calendar isn’t connected, start instead with: “Here’s my day: [paste your meetings]. Build me a morning brief from this plus my inbox.”)

Read what comes back. Tweak it. Too long? Say “shorter”. Missing something? Add it. Get it to the version you’d actually want first thing.

Step 4.
Save it as a job you can run with one tap. This is the version that works for everyone, today, no setup. Paste this to Claude:

Save that exact brief as a reusable prompt I can run each morning.
Write it into a file called morning-brief.md in my Switched-On folder, with clear instructions.
Then each day I can open a fresh chat, point at this file and say “run my morning brief”.

Now every morning is one line: “run my morning brief”. The job is built. You just press go.

Step 5.
Optional upgrade, if your app can schedule it. Some setups let you schedule a task so it runs on its own, before you’ve sat down. If yours does, set the morning brief to run early each weekday and land where you’ll see it. If it doesn’t, or you’re not sure, don’t worry. The one-tap version from Step 4 gives you almost all of the benefit with none of the fuss.

Two rules if you do schedule it. Keep the draft-only guardrail in the instructions, word for word. And learn how to switch it off before you switch it on, see the safety note.

Step 6.
Lock it in. Paste this to Claude:

Add to MEMORY.md: Day 20, built my first agent, a morning brief.
It reads my day and drafts a summary. It never sends or replies.
I run it with one tap, or on a schedule if set up.

A safety note

An agent that can act on your behalf is powerful, so keep the leash short, especially at the start.

DRAFT ONLY. Your morning brief reads and summarises. It does not reply, send, accept, decline or delete anything. Those words stay in the instructions. The day it starts doing things on its own is the day you lose track of what it’s doing in your name.

If you schedule anything, know how to stop it first. Find where scheduled or recurring tasks live in your app and check you can pause or delete one. Never start a job you don’t know how to switch off.

What you’ll notice

The first morning the brief is just there, you’ll feel the difference. No scramble. No five tabs. The day, already gathered, while you found your shoes.

You’ll also notice the quiet shift in how you think about AI. Up to now you’ve asked it for things. Now it’s started doing a small job for you, on repeat. That’s the door to the last stretch of this challenge.

Why this works

The morning scramble is small but it’s daily, and daily adds up. It’s also entirely made of gathering, checking the calendar, scanning the inbox, recalling the priorities, which is exactly the kind of fetch-and-summarise work AI does well and you do grudgingly.

Building it by hand first is the trick. You see exactly what it does, you shape it, and only then do you trust it to repeat. Trust is earned by watching, not by hoping.

Key takeaway

An agent is just a job you set up once and get back done. Five parts: trigger, job, context, output, guardrail.

Your first one is built. The morning, gathered for you. Draft only, on a short leash, switched on by one tap or on a schedule. You stay the boss.

Where this is heading

Friday tomorrow, end of Week 3. We slow down and put the whole brain to work on the third and sharpest way to use AI. The one we named on Day 16 but saved. AI as the one who argues back.

Tomorrow on Switched On

Day 21: Let Claude argue against you.

The decision you’ve been circling for three weeks. Tomorrow you settle it, with an honest opponent in the room.

Nishma x

P.S. How did today land? Three buttons, thirty seconds. Tap here

I read every response. The data shapes the rest of the challenge.

Catching up? Days 0 to 19 are on the Switched On Substack archive HERE.

Got a friend or a colleague who this might be useful for? Send her to nishma.co/30daychallenge

#SwitchedOn #SwitchedOnAIChallenge

​Unsubscribe | Update your profile | 167-169 Great Portland Street, London, W1W 5PF

Welcome to Day 19 of the challenge.

Some of us run our business on a feeling and a glance at the bank balance and a speed-dial to our accountant.

We know roughly what’s coming in. We know it’s been a good month or a tight one. We can feel when something’s off. What we often don’t have is a clear picture of where the money actually comes from, where it quietly leaks, and what next month really looks like if nothing changes.

If you’re not at the stage of having a fantastic CFO, and you don’t have a finance degree, you can still show Claude (ChatGPT, Gemini, etc.) your numbers and ask the right questions.

Today you don’t just talk about the money. You build something that does the maths and shows you the picture.

A word before we start

This is where Claude stops being a writer and becomes an analyst.

Give it a pile of numbers and a clear question, and it will find the pattern, build the model and explain what it sees in plain English. The same thing a good finance person does, except it doesn’t sigh when you ask it to redo it a fourth time.

One honest line before we start. What Claude builds today is a model and a read on your numbers. It is not regulated financial advice, and it isn’t a substitute for your accountant. Use it to see clearly and ask better questions. Sense-check the big decisions with a professional. With that said, let’s build.

Step by step

Step 1. Pick your track. Choose the one that matches what’s actually on your mind.

The cash-flow forecaster. Best if the question keeping you up is “will there be enough?” Claude builds a simple month-by-month forecast of money in and money out.

The P&L read. Best if you have a profit and loss statement but never really interrogate it. Claude reads it and tells you where margin is eroding, where costs have crept and how dependent you are on one client.

The pipeline analyser. Best if you sell to clients and carry deals. Claude turns your list of live opportunities into a weighted forecast and shows you where deals stall.

Pick one. You can do the others another week.

Step 2. Gather your numbers. Keep it rough, Claude will tidy it up.

1. Cash flow: your regular money in (clients, retainers, products) and money out (salaries, rent, subscriptions, tax set-aside), with rough amounts and how often.

2. P&L: your most recent profit and loss, exported as a spreadsheet or copied as text.

3. Pipeline: your live deals, each with a rough value, the stage it’s at and how likely it is to close.

Step 3. Open Cowork in your desktop app, pointed at your Switched-On folder. Opus for this one.

Step 4. Brief it. Here is the cash-flow version. Adapt the first lines for whichever track you picked. Paste this to Claude:

I want to build a simple twelve-month cash-flow model for my business. Here are my regular money-in and money-out items with rough amounts and frequency: [paste]. Ask me anything you need to fill the gaps, one question at a time. Then build me a clear month-by-month model as a spreadsheet, show me the running balance, and tell me in plain English the three things this reveals, including any month that looks tight.

For the P&L read, paste this to Claude:

Here is my latest profit and loss [paste or attach the file]. Read it like a sharp finance director who’s on my side. Tell me where my margin is being eaten, where costs have crept up, how reliant I am on any single client or product, and the three questions I should be asking that I’m probably not.

For the pipeline, paste this to Claude:

Here are my live deals, each with rough value, stage and likelihood: [paste]. Build me a weighted forecast, show me total committed versus best-case, tell me where deals are stalling, and which two I should chase this week to move the number most.

Step 5. Make it something you’ll actually look at. A spreadsheet you never open is no better than a feeling. So, paste this to Claude:

Now build me a single-page dashboard as an HTML file I can keep open. Big clear numbers at the top, one simple chart, and a short what this means line under it. Clean, uncluttered. Save it as money-dashboard.html in my Switched-On folder.

Step 6. Capture the honest summary. The by-product of today is a clear-eyed note on the business itself. Save it. Paste this to Claude:

Based on everything we’ve just done, write me a short, honest business.md: what I sell, who actually pays me, what’s working, what’s leaking, and what I’m quietly worried about. Plain and direct, the version I’d never put in my marketing. Save it at the top of my Switched-On folder and reference it in CLAUDE.md.

Then, paste this to Claude:

Add to MEMORY.md: Day 19, Claude built a cash-flow model (or P&L read, or pipeline forecast) and a dashboard from my real numbers. business.md now holds the honest version.

A safety note

This is the most sensitive file you’ve made so far. Treat it that way.

NEVER paste in bank account numbers, card numbers, sort codes or logins. Claude needs amounts and categories, not credentials. Round numbers are fine.

These files live on your laptop, in your folder. business.md and the dashboard are for your eyes. Don’t email them around or drop them in a shared drive without thinking twice.

What you’ll notice

The leak you half-knew about, now in black and white. The client who feels big but isn’t, or the small one quietly carrying you. The month that’s going to be tight, far enough ahead that you can do something about it.

And the relief of seeing it. Money worry thrives in the fog. A clear model doesn’t make the numbers bigger, but it makes them knowable, and knowable is calmer.

Why this works

You’ve been carrying the maths in your head, which is exhausting and inexact. Claude does the calculation without judgement and explains it without jargon. You keep the decisions. It does the arithmetic and the picture.

This is also the file that quietly upgrades everything else. Once Claude knows how your business actually makes money, its help on pricing, offers, even content gets sharper, because it knows what you’re really trying to grow.

Key takeaway

You don’t have to be a numbers person to see your numbers clearly.

A pile of figures in. A working model, a dashboard and an honest business.md out. The money side of the business, finally out of the fog.

Where this is heading

Tomorrow we hand Claude its first proper job to do on a schedule. The morning catch-up that puts itself together before you’ve had your tea.

Tomorrow on Switched On

Day 20: Build the morning brief that runs itself.

Your first agent. A job you hand over once and get back every day.

Nishma x

P.S. How did today land? Three buttons, thirty seconds. Tap ​here​

Catching up? Days 0 to 18 are on the Switched On Substack archive ​HERE​.

Got a friend or a colleague who this might be useful for? Send her to nishma.co/30daychallenge

SwitchedOn #SwitchedOnAIChallenge

Welcome to Day 18 of the challenge.

Most audience work is fiction.

You sit down, invent a woman, give her an age and a job and a fondness for a glass of wine on a Thursday, call her a persona and write to her forever. The trouble is she doesn’t exist. So everything written to her lands as generic, because generic is what you get when you write to a made-up average.

You don’t need to invent her. You’ve already met her.

She’s in your inbox. Your DMs. Your comments. The replies to your last newsletter. The people who actually buy from you, read you and write to you are telling you who they are, in their own words, every week. Today Claude reads the evidence and finds the pattern.

One honest thing first. This isn’t a replacement for proper audience research. If you’ve done the deep work, segmentation, interviews, the real thing, keep it. What today gives you is a sharp, evidence-based brief you can hand any AI so that everything it helps you make for your audience lands closer to home. It sits alongside the proper work. It doesn’t pretend to be it.

A word before we start

The reason this works is the same reason the voice profile worked on Sunday. You don’t tell AI who your audience is. You show it the evidence and let it find the pattern.

A description of your audience is a guess dressed up as a fact. The actual words real people use when they write to you are not a guess. They’re data. Claude is very good at reading a pile of real messages and telling you what keeps coming up: the same worry, the same phrase, the same question asked five different ways.

Step by step

Step 1. Gather the real evidence. Pick whichever of these you can get to easily:

1. Twenty or thirty real messages people have sent you. DMs, comment replies, emails from readers or clients. Copy the text into one file. Strip out surnames and anything identifying as you go.

2. Your email, if you’ve connected Gmail (Day 7). Claude can read the recent ones for you, no copying needed.

3. A list of clients or subscribers if you keep one, even a rough spreadsheet.

Don’t polish or cherry-pick the nice ones. The annoyed message and the confused one are often worth more than the gushing one.

If you’re pasting messages in, save them to your reference folder as audience-evidence.md.

Step 2. Open Cowork in your desktop app, pointed at your Switched-On folder. Opus for this one. Make sure CLAUDE.md is at the top.

Step 3. If you’re using pasted messages, paste this to Claude:

Read the file reference/audience-evidence.md. These are real messages from people in my audience. Find the pattern. Tell me: who keeps showing up, what they actually ask for, the exact words and phrases they use, the worry underneath the question, and what they never say but clearly feel. Don’t smooth their language into marketing speak. Quote them where it’s telling. Ask any clarifying questions before you begin.

If you connected Gmail and want Claude to read your inbox instead, add this line at the start:

Read the last fifty emails I’ve received from people who aren’t colleagues or suppliers. Treat them as audience evidence. And... (then add the rest of the prompt above)

Step 4. Read what comes back.

This is the bit that lands. You’ll see your audience described in their own words, not yours. The question you didn’t realise everyone was asking. The phrase three different people used. The thing they’re all quietly worried about.

Step 5. Now sharpen it on one real person. Patterns can still drift into a blur. So pin it to someone real. Paste this to Claude:

Now pick the one real person from this evidence who most represents the pattern, or let me name her. I’ll tell you about her. Then write me audience.md: who she is, what she actually does, what she’d say at dinner that she’d never post, what she’s scrolling past, and what she genuinely wants from me. Use her real words from the evidence. First name only, nothing that identifies her.

Answer any follow-ups honestly. The uncomfortable specifics are the useful ones.

Step 6. Save it into the spine. Paste this to Claude:

Save the final file as audience.md at the top of my Switched-On folder. Then update CLAUDE.md with a line: See audience.md for who I actually serve, built from real evidence.

Step 7. Optional, build her a card you can keep open while you write. Paste this to Claude:

Take audience.md and make a single-page HTML card I can keep open in my browser. Clean, dark background, generous spacing. Five blocks: who she is, what she’s really asking for, what she’d never admit, what she scrolls past, what she wants from me. Save it as audience-card.html.

A safety note

These are real people. Treat the file with care.

FIRST NAMES ONLY. No surnames, no employer, no address, nothing that would identify someone if the file ever leaked. Describe her role, not her organisation. If the person who best fits is genuinely private, someone who confides in you, use a made-up name. Claude is using the pattern, not the identity.

If you used Gmail, the evidence stays in your session. Don’t save anyone’s raw emails into a file you’d share.

What you’ll notice

She comes to mind every time you sit down to write now. Not a demographic. A person, with a real worry and a real way of talking.

You’ll also notice your own language change. Your sentences get shorter. Your examples get specific. The jargon falls away, because she’d roll her eyes at it. That’s the whole point. Write to one real person and the others like her lean in.

Why this works

Real beats representative. A persona is an average, and nobody is an average. The actual words of twenty real people carry the texture an invented profile never will, the specific fear, the specific phrase, the specific thing they’ve tried and given up on.

You had the evidence all along. You’d just maybe never sat down and read it as ‘evidence’. Claude does the reading and finds what repeats.

Key takeaway

You don’t invent your audience. You read it back from the people already there.

Real messages in. A pattern out. One real woman to sharpen it on. One audience.md at the top of your folder, built from evidence, not a vibe.

Where this is heading

Tomorrow, the money side. And we don’t just write about it, we build something that does the maths and helps us with key business decisions.

Tomorrow on Switched On

Day 19: Build the money model your business actually needs.

No finance degree required. Show Claude the numbers, ask the right question, and walk away with a working model you’ll actually use.

Nishma x

P.S. How did today land? Three buttons, thirty seconds. Tap ​here​

Catching up? Days 0 to 17 are on the Switched On Substack archive ​HERE​.

Got a friend or a colleague who this might be useful for? Send her to nishma.co/30daychallenge

SwitchedOn #SwitchedOnAIChallenge

​

Welcome to Day 17 of the challenge.

A small change of plan first. Yesterday I said today was the audience day. It’s not. The audience work lands tomorrow, and it’s better for the wait. Today we do something you’ll feel in your shoulders.

You have a folder you don’t open.

We all do. The Downloads folder with four years of screenshots and a file called final_v3_ACTUAL.docx. The Desktop you’ve stopped seeing. The client folder that became a dumping ground. It’s the digital version of the drawer in the kitchen with the batteries and the spare keys and a single chopstick.

Today Claude opens it. And tidies it.

Not a list of tips on how you should tidy it one day. Claude actually does the work, in front of you, on your machine.

​

A word before we start

Up to now, you’ve mostly used Claude as something you talk to. Ask a question, get an answer. Useful, but a sliver of what it can do.

This is the day that changes. Claude can read the files on your computer, see what’s there, sort it, rename it and file it away. It’s not a chatbot in a box. It’s a pair of hands.

That can feel like a big step. So you get to choose your nerve. There are two ways to do today, and neither is wrong.

The cautious way: Claude proposes everything and asks your permission before it moves a single file. You watch. You approve. Nothing happens without your yes.

The brave way: Claude tidies the whole thing in one go and shows you exactly what it did afterwards, with a list so you can undo anything you don’t like.

Most people will pick cautious. The brave ones feel the magic faster. Pick the one that lets you sleep.

​

Step by step

Step 1.

Pick the folder.

Downloads is the one everyone has and everyone dreads. Start there. If you’re feeling braver, pick a real work folder, the client one that turned into a swamp.

One rule before you start: this should not be the only place an irreplaceable file lives. If there’s something in there you could never get back, copy it somewhere safe first, or pick a different folder.

Step 2.

Open Cowork in your desktop app. Point it at the folder you chose, not your Switched-On folder today. Opus for this one.

(Reminder: the toggle at the top of the app reads Chat, Cowork and Code. You want Cowork. Ignore Code, that’s the developer side.)

Step 3.

First, just look. Paste this to Claude:

“Look in this folder. Don’t move or rename anything yet. Tell me what’s actually in here: what types of files, roughly how many, what’s old, what looks like junk, what looks important, and what natural groups you can see. Give it to me as a short summary.”

Read what comes back. This alone is worth the day. Most people have no idea what’s actually in there.

Step 4.

Now choose your nerve.

The cautious version, paste this to Claude:

“Propose a folder structure to tidy this. Show me the groups and where things would go. Then move files one batch at a time, and ask me to confirm before each batch. Don’t rename or bin anything without telling me first.”

The brave version, paste this to Claude:

“Tidy this folder. Create sensible subfolders, file everything into them, and flag anything that looks like junk or a duplicate in a folder called Review me rather than deleting it. When you’re done, give me a list of every change you made so I can undo anything I don’t like.”

Notice the safety in the brave version. Nothing gets deleted. Suspected junk goes to Review me, where you make the final call. Claude flags. You bin.

Step 5.

Build the map. Once it’s tidy, ask for an index so future-you can find things. Paste this to Claude:

“Create a file called INDEX.md at the top of this folder. List each subfolder and one line on what’s in it. This is the map I’ll read when I can’t find something.”

Step 6.

Lock in what you learned. Paste this to Claude:

“Add this to MEMORY.md: Day 17, Claude can tidy and file a real folder, not just talk about it. It moves, renames and indexes. Junk goes to Review me, never straight to the bin.”

​

A safety note

This is the first day Claude touches your actual files, so read this twice.

NEVER point Claude at the only copy of something you can’t lose. Copy it somewhere safe first, or use the cautious version where you approve every move.

Nothing today should be deleted outright. Suspected junk and duplicates go to a Review me folder. You make the final call on the bin. If Claude ever offers to delete, say no and ask it to move to Review me instead.

Start with Downloads, not your most important client folder. Build trust on the low-stakes mess first.
​

What you’ll notice

The summary in Step 3 is the quiet shock. You’ll see how much you’ve been carrying without knowing. Hundreds of files you forgot existed. Three versions of the same document. A folder’s worth of screenshots you never looked at again.

Then the tidy happens and your shoulders drop. The thing you’d been meaning to do for ages, the job that always lost to something more urgent, done while you watched.

And the bigger thing: you’ll stop thinking of Claude as a clever writer and start thinking of it as someone who can actually do jobs.
​

Why this works

Tidying is boring, fiddly and exactly the kind of work that never reaches the top of the list. It’s also pattern-matching and sorting, which is what AI is genuinely good at. The match is perfect. You bring the judgement about what matters. Claude brings the patience to sort two thousand files without getting bored.

You kept the control where it counts: nothing deleted, every change logged, your yes required where you wanted it.
​

Key takeaway

Claude doesn’t just write. It does jobs. Today it tidied a real folder, filed it, and drew you a map.

The folder you’ve been avoiding for ages. Sorted. And you learned the thing that unlocks the rest of this week: Claude works on your stuff, not just in a chat window.
​

Where this is heading

Tomorrow, the audience work I promised you. The same move you saw today, give Claude the real thing instead of a guess, pointed at the people you actually serve.
​

Tomorrow on Switched On

Day 18: Stop guessing who you’re talking to.

The people who actually buy from you, read you and reply to you are already telling you who they are. Tomorrow Claude finds the pattern.

​

Nishma x

​

P.S. How did today land? Three buttons, thirty seconds. Tap here​

​

Got a friend or a colleague who this might be useful for? Send her to nishma.co/30daychallenge

SwitchedOn #SwitchedOnAIChallenge

Welcome to Day 16 of the challenge.

Yesterday Claude got to know who you are. Today Claude learns how you sound.

Today is also bigger than that. It’s the day we draw the line that protects everything from here on.

There’s a question every woman doing serious work with AI has to answer: when does AI get to write FOR me, and when does it only get to help me write? Most AI training skips this question. Most AI users default to “AI writes everything” and produce slop in their own name. Some go the other way and never use AI for any writing, leaving hours on the table.

Today we build the voice profile. AND we draw the line. The line is the more important teaching.

A word before we start

Voice is not an adjective. It’s a pattern.

How you open. How you close. The rhythm of your sentences. The vocabulary you reach for. The vocabulary you avoid. Where you’re specific. Where you’re general. Where you let a sentence run. Where you stop dead.

Most people asked to describe their voice write a list of adjectives (”warm, direct, no jargon”) and call it done. Claude can’t write to adjectives. Claude CAN read a pattern across ten examples and triangulate the real voice underneath.

Same move as Day 5: show, don’t tell.

Once we have the pattern in a file, the question becomes: when do we want Claude using it?

The Switched On line

Here’s the rule we’re going to hold for the rest of this challenge, and the rest of your AI use after.

Three modes. One bar.

AI as editor. When your voice IS the product.

LinkedIn posts. Substack essays. Instagram captions where personal brand is the point. Keynote scripts. Thought leadership. Anything published under your name where the writing IS the value.

In these moments, YOU author the words. Claude edits. Claude sharpens. Claude flags what’s lazy and what’s buried. You publish your own words, made sharper.

AI as author. When the work IS the product.

Proposals. Emails. Client follow-ups. Marketing collateral. Internal briefs. Decks. Operational comms. Anything where the THING being delivered is the work, not your voice on a page.

In these moments, Claude can draft. You sharpen and send. The voice profile means the AI-authored draft still sounds like you. But the writing isn’t the point. Getting the work done is. Important. AI drafting is permission to start. Not permission to ship. The quality bar doesn’t drop because something is “ops.” Sharpen the AI draft like you wrote it yourself. The minute you let “ops” mean “good enough,” AI slop creeps into your brand by the back door, one email at a time.

AI as challenger. When you need adversary thinking.

The decision you’ve been weighing. The post you’re not sure about. The argument you can’t quite land. Claude plays the difficult colleague your inner circle is too polite to be. AI argues against you. You decide.

This is the most powerful of the three modes for senior women in leadership. We come back to it properly on Day 22. Today we lay the foundation by building the voice profile that makes the adversary thinking land sharper.

The voice profile we’re building today serves all three modes differently:

• For editor work, it makes Claude a sharper EDITOR - using your voice patterns as the editing standard.

• For author work, it makes Claude a sharper GHOSTWRITER on the ops stuff - drafts arrive sounding like you, requiring less rewriting.

• For challenger work, it gives Claude context for your actual defaults and blind spots, so the adversary thinking lands on YOU, not on a generic version of you. We demo this on Day 22.

One file. Three uses. One quality bar across all of them. The line is yours to hold.

A note on why this matters: the cohort members who get most from AI long-term are the ones who hold this line. The ones who don’t (who let AI write everything) produce content that sounds increasingly generic, lose trust with their audience, and end up paying more to dig out of the AI-slop hole than they ever saved.

You’re building the alternative. The version where AI gives you leverage on the work without taking your voice away.

Step by step

Part 1: Build the voice profile

Step 1. Gather ten pieces of writing you’re proud of.

Not the polished website copy. Real things you’ve written. LinkedIn posts that got reactions. A Substack essay you’d send to your mum. An email that landed. A section from a keynote. A long DM reply. Anything where you wrote it and thought “yeah, that’s me.”

Pick on the basis of bravery, not popularity. This is the one rule that matters today. The voice profile compounds whatever you feed it. If you train it on the safe stuff (the posts that got the most likes because they were inoffensive), you get safer AI output. If you train it on the brave stuff (the posts you nearly didn’t publish, the ones with the cleaner edge), you get bolder AI output. Most cohort members make this mistake. Don’t.

Ten is the sweet spot. Fewer and Claude over-fits to one mood. More and you lose specificity.

Paste them into a single text file. Label them lightly: “LinkedIn post about X (2025)” or “Email to client Y”. Just enough that Claude knows the format. Save to your reference/ folder as voice-samples.md.

Step 2. Open Cowork in your desktop app, pointed at your Switched-On folder. Opus for this one. Make sure CLAUDE.md exists at the top (Day 14 work).

Step 3. Paste this brief:

Read the file reference/voice-samples.md. These are ten pieces of writing I’m proud of.

Write me a voice profile. Cover, specifically:

1. How I open. Pattern, length, what I do in the first sentence. 2. How I close. Pattern, length, what I do in the final line. 3. Sentence rhythm. Short, long, mixed, where I let one run, where I stop dead. 4. Vocabulary I reach for. Words and phrases I use often. 5. Vocabulary I avoid. Words you can see I’d never use. 6. Where I’m specific. Numbers, names, places, examples. 7. Where I’m general. The territory I leave deliberately vague. 8. Where I’m vulnerable. The cracks I let in. 9. Where I’m confident. The places I plant a flag. 10. The tells that would make a friend say “that sounds like Nishma.”

Be ruthless. Don’t smooth me. Don’t sanitise. The non-obvious patterns matter more than the polite ones.

Save the output to the top of my Switched-On folder as voice-profile.md. Once saved, also update CLAUDE.md to reference it.

Step 4. Read the draft.

Two checks:

• Does it match what you’d say if asked to describe your voice?

• Does it include the non-obvious tells (the swearing, the structural quirk, the words you reach for that aren’t on the polished website)?

If yes, approve and save. If no, tell Claude what’s wrong. “You missed that I almost never use exclamation marks. I use full stops as percussion. Add that.”

Part 2: Test it in editor mode (voice work)

Now we test the voice profile against the line. First in editor mode - where YOU author and Claude edits.

Step 5. Pick a half-written LinkedIn post you’ve got lying around. Notes from a thought you’ve been meaning to publish. A rough draft. Voice notes you transcribed. Anything 100-200 words of your own writing that isn’t quite there yet.

If you don’t have one, take five minutes (Wispr Flow) and dictate one now. Real thought, half-formed.

Step 6. Start a fresh chat in your Switched-On folder. DON’T load the voice profile yet. Paste:

Here’s a half-written LinkedIn post. Edit it. Sharpen the hook, cut what’s flabby, strengthen the closing. Don’t rewrite. Edit. Keep my words where they work.

[Paste your draft.]

Read the edit. Note what Claude did.

Step 7. Start ANOTHER fresh chat. This time, make sure voice-profile.md gets read. Paste:

Read voice-profile.md before doing anything else. Then here’s a half-written LinkedIn post. Edit it the way I would edit it - using my voice patterns as the editing standard. Sharpen the hook, cut what’s flabby, strengthen the closing. Don’t rewrite. Keep my words where they work.

[Paste the same draft.]

Read this edit. Compare to the first one.

Step 8. Read both edits side by side.

The first edit will be technically competent. The second edit uses the same skills, but the cuts and suggestions land closer to what YOU would have cut yourself. The voice profile didn’t make Claude a better writer. It made Claude a better editor OF YOUR work.

This is editor mode. The thing you’ll do every time you publish something under your name. Your words, sharper.

Part 3: Test it in author mode (operational work)

Now we shift sides, where Claude authors and you sharpen.

Step 9. Pick a real operational task you need done. Examples:

• A follow-up email after a meeting this week

• The opening paragraph of a proposal you’re drafting

• A LinkedIn DM you need to send to a connection

• A short brief for a colleague

• A holding email to a client who’s chasing

Something where the WORK is the deliverable, not your voice on a page.

Step 10. Start a fresh chat. DON’T load the voice profile yet. Paste your context and ask Claude to draft.

I need to send a follow-up email after a client meeting. Context: [3-4 lines about the meeting, what was agreed, what needs to land in the email]. Draft it.

Read the draft. Note how generic it sounds.

Step 11. Start another fresh chat. Load voice-profile.md. Paste the same brief.

Read voice-profile.md before doing anything else. Same task: a follow-up email after a client meeting. Context: [3-4 lines]. Draft it in my voice.

Read this draft. It will be much closer to something you’d actually send. Sharpen the last 10%, send it, done in five minutes instead of thirty.

This is author mode. The thing you’ll do for proposals, emails, marketing collateral, internal docs. The operational writing where AI gives you back hours without taking your voice away.

Step 12. Save the comparisons and lock in the lesson.

Ask Claude:

Save the four comparison drafts to outputs/2026-XX-XX_voice-profile-test/ The two editor-mode edits (with and without voice profile) and the two author-mode drafts (with and without voice profile). These are my evidence of what voice-profile.md actually does.

Then update MEMORY.md:

Day 16: built voice-profile.md and tested it in both editor and author mode. The rule from here: Voice work - I write, Claude edits. Operational work - Claude drafts, I sharpen. The voice profile makes Claude better at both, in different ways.

A safety note

The voice profile is yours. Treat it as part of your brand asset library.

If you ever change roles, rebrand, or radically shift register (your post-corporate voice will differ from your corporate one), rerun the voice profile from 10 new examples. Your voice evolves; the file should too. Once a year is enough for most.

On the framework itself: the line you draw across all three modes is the safeguard against AI slop. Hold it. The day you start letting AI author your social posts is the day your audience starts to feel the difference, even if they can’t name it.

What you’ll notice

The editor demo is where your protection lives. AI editing your voice work is faster and sharper than editing it alone, but the work is still yours. You publish your own words.

The author demo is where the hours come from. AI drafting your proposals and emails saves you the blank-page friction on the work that doesn’t need to be a piece of YOU. The voice profile means those drafts arrive sounding like you, requiring minutes of editing rather than hours.

Most cohort members feel relief at the framework before they finish reading this. The relief is the answer to a question you’ve been sitting on: “How much should I let AI write?” The answer is now specific. Depends what kind of writing it is. Use the line.

Why this works

Voice is the pattern across many pieces, not the description of one. The traditional brief (”write in my voice, which is warm and direct”) fails because warm-and-direct describes a thousand people. The pattern-from-examples brief works because it captures what’s actually only true of you.

That makes the voice profile useful in both modes. As an editing standard (editor mode), the file gives Claude a benchmark for what “good” looks like. As an authoring style (author mode), the file gives Claude a starting voice that’s closer to yours from the first draft.

The line itself, when to use which mode, is the more durable teaching. AI is changing every six months. The author/editor line will be the same in 2030 as in 2026.

Key takeaway

Voice profile saved to the spine. Three modes named. Two demoed today, the third coming on Day 22.

AI as editor on the voice work: posts, essays, anything published under your name. You write. AI sharpens.

AI as author on the operational work: proposals, emails, marketing collateral, internal docs. AI drafts. You sharpen and ship to your own bar.

AI as challenger on the thinking. When you need adversary perspective. Demo on Day 22.

Three modes. One quality bar. The line is yours. Hold it.

Where this is heading

You’ve now got two brain files at the top of your Switched-On folder: about-me.md (who you are) and voice-profile.md (how you sound). Both read on every session. Both compounding.

Tomorrow is the third. The most consequential for marketing.

Tomorrow on Switched On

Day 17: Stop writing to demographics. Write to her.

Pick one real woman in your audience. Not a persona. A real person. Tomorrow Claude interviews you about her. The file that comes out changes how you write to anyone, across all them.

Nishma x

P.S. How did today land? Three buttons, thirty seconds → ​Tap one​

Catching up? Days 0 to 15 are on the Switched On Substack archive ​HERE​.

Got a friend or a colleague who this might be useful for? Send her to nishma.co/30daychallenge

SwitchedOn #SwitchedOnAIChallenge

Welcome to Day 15. Saturday. Week 3 of the challenge begins.

The first two weeks were about giving AI work to do. Connect Gmail. Triage the inbox. Prep the meeting. Read the doc. Kill the Sunday Scaries. Build the dashboard. The pattern was: you have a task, Claude does the assembly, you make the call.

This week is different.

This week we tell Claude who YOU are, properly. So that when Claude does the assembly, the assembly already knows your standards. Your voice. Your audience. Your business. What you’d never do. What you’re known for. The non-obvious things about how you actually work.

It starts today with a Saturday-morning interview. Ten minutes. Claude asks. You answer. By the end, your first proper brain file.

A word before we start

Don’t write this yourself. That’s the mistake.

If you sit down to write “about me” you’ll second-guess every sentence. You’ll write the polished version. You’ll smooth the edges. The Claude that reads it later will get a corporate bio, not you.

The interview is the move. Claude asks one question at a time. You answer. You don’t see the questions coming. You don’t have time to over-shape your answers. The file that comes out is sharper, weirder and more useful than anything you’d have written from a blank page.

​Wispr Flow​ earns its keep today. Talk your answers instead of typing them. The voice answers are always better than the typed ones.

A note for those of you still typing: it’s fine. Slower, less honest, still works. The format is the same.

Step by step

Step 1. Open Cowork in your desktop app, pointed at your Switched-On folder. Opus for this one.

Step 2. Paste this brief:

You’re going to interview me for about ten minutes to build a detailed about-me.md document. The file lives at the top of my Switched-On folder and will be referenced by you and any future AI as the primary brief on who I am professionally.

Ask me one question at a time. Wait for my answer. Don’t move on until I’ve given you a real answer (push back if my answer is vague). Cover:

1. What I actually do (not the job title, the work). 2. Who I serve and who I deliberately don’t. 3. What makes my work different from others doing similar things. 4. What I’m building toward over the next 12-24 months. 5. What I refuse to do. The work I turn down or won’t take. 6. The non-obvious things about how I work (pace, hours, formats I love and hate, who I bring in, when I’m sharpest, when I’m not).

Start with question one.

Step 3. Answer. One at a time. Honestly.

If Claude asks a follow-up, answer the follow-up. If the question feels uncomfortable, that’s the right question. Tell it more, not less.

Step 4. After question six (and any follow-ups), paste this:

Now draft about-me.md based on my answers. Use my words where I gave you good ones. Don’t smooth me into a corporate bio. Keep the specifics. The non-obvious things matter most.

Save to the top of the Switched-On folder as about-me.md once I’ve approved.

Step 5. Read the draft.

Two checks:

• Does it sound like you, not like a LinkedIn summary?

• Does it include the non-obvious things you told it?

If yes, approve and save. If no, tell Claude what’s wrong and ask for another pass.

Step 6. Now link it into the spine.

Update CLAUDE.md to reference about-me.md. Add a line under “Who I am” that points to the new file: “See about-me.md for the full brief on me.”

Step 7. Add to MEMORY.md.

Add this to MEMORY.md: “Day 15: built about-me.md from a Saturday interview. The non-obvious things are the most important. Don’t smooth them.”

Save. Close.

What you’ll notice

The answers are sharper than your bio.

The questions Claude asks aren’t the questions you’d put on your website. “What I refuse to do” doesn’t make it into copy, but it’s the most useful sentence in the file. “The non-obvious things about how I work” is the section every future Claude session will lean on most.

You’ll also notice you’ve never said some of these things out loud before. That’s the point. The brain file is the first time some of it gets recorded.

Why this works

Voice and self-articulation are different jobs. Voice you can train on examples (we do that tomorrow). Self-articulation has to come from you.

But you can’t get the articulation by sitting at a blank page. The blank page invites the polished version. The interview catches the unpolished version, which is the useful one.

This is why interviewers exist as a profession. Most people are interesting under the right questions and boring at the keyboard. Today Claude is the interviewer.

Key takeaway

Ten minutes. Six questions. Answered honestly. One brain file at the top of your Switched-On folder.

From now on, every Claude session in this folder reads it. Every brief you write gets shorter. Every output gets sharper. Because Claude already knows.

Going deeper, when you have a Sunday afternoon

The about-me.md you’ve just built is the working version. There’s a deeper version of this exercise that’s worth doing once in your life, ideally in the next month.

Ruben Hassid, one of the sharpest AI educators out there, has published the full methodology in a piece called I Am Just A Text File. It’s a 100-question interview across seven categories — Beliefs & Contrarian Takes, Writing Mechanics, Aesthetic Crimes, Voice & Personality, Structural Preferences, Hard NOs, Red Flags. About two hours, run with Claude Opus. The output is the deepest brief any AI will ever have about you.

Read his post and run the exercise on yourself: ​ruben.substack.com/p/i-am-just-a-text-file​

Save the output to your Switched-On folder as about-me-deep.md. From then on, brief any AI with that file.

His framing is worth reading even before you run the exercise. “Master yourself so AI can master you” is the line that stays with you.

The 10-question version you built today gets you 80% there in 1/10th the time. The 100-question version gets you 100% there. Both are valid. Do the deep one when you’ve got the bandwidth.

Where this is heading

This is the first of six brain files going in across Week 3. Voice tomorrow. Audience after. Then pillars, business, week-pattern. By next Friday, the brain about you is in.

By the end of next week, you’ll write a LinkedIn post in 10 minutes that takes others 90.

Tomorrow on Switched On

Day 16: Make AI write like you on a good day.

Sunday edition. Ten things you’ve written. One prompt. A voice profile any AI can read for years.

Nishma x

P.S. How did today land? Three buttons, thirty seconds → ​Tap one​

I read every response. The data shapes the rest of the challenge.

Catching up? Days 0 to 14 are on the Switched On Substack archive ​HERE​.

Got a friend or a colleague who this might be useful for? Send her to nishma.co/30daychallenge

SwitchedOn #SwitchedOnAIChallenge

Welcome to Day 14 of the challenge.

This is the most important day of the challenge so far.

It’s also one of the days I’ve been most nervous about you reading. Not because it’s hard. It isn’t. Twenty minutes of doing, ten minutes of reading. But because everything so far has been pitched at “AI is now your colleague” and today is the day we cross into “AI is now your colleague who knows you.”

Two text files. One paragraph pasted into a settings menu. That’s it.

From tomorrow on, every Claude session you open inside your Switched-On folder reads those two files first and behaves as if it already knows you. It does, because you’ll have told it.

A word before we start

Two reassurances before the technical bits.

One: this is not technical. The files are .md text files. The contents are plain English. The setting is a paragraph in a settings menu. If you can write a paragraph and copy-paste a sentence, you can do today’s task.

Two: it persists. From tomorrow, Claude opens your folder and reads these files automatically. You don’t have to remember to do anything. You don’t have to upload anything. You set this up once and the benefit compounds across every chat you ever have.

It looks like infrastructure because it persists. It’s actually just text files in a folder.

Why .md? Why these specific filenames?

Quick aside before we begin.

The files we’re creating end in .md, which stands for markdown. It’s a plain text format that lets you add light structure (headings, bold, bullets) without locking the file into Word or any specific app.

Three reasons it matters:

1. Future-proof. Any AI can read markdown. Any text editor can open it. Ten years from now, the brain you build today still works.

2. Lightweight. No formatting bloat. No fonts. No corruption. Just words and a few asterisks for emphasis.

3. AI-native. AI models are trained heavily on markdown. They read it better than Word documents or PDFs.

The specific filenames matter too. CLAUDE.md is a convention Claude is trained to look for. Put a file called CLAUDE.md at the top of a folder, and Claude reads it automatically every time it opens that folder. Same for MEMORY.md.

That’s literally the whole magic. Name the file correctly, put it in the folder, Claude does the rest.

What this folder will become

Right now your Switched-On folder has the three sub-folders from yesterday plus a handful of outputs from the past two weeks. By Day 30 it’ll have:

• Eight foundational files that capture who you are, how you sound, who you serve, what you sell and what you’re known for

• A working agent skill that runs on a schedule

• A folder of dozens of outputs from work Claude has done for you

• A running log of decisions you’ve made with Claude as your thinking partner

The two files we create today (CLAUDE.md and MEMORY.md) are what makes everything else possible.

Twenty minutes well spent.

Step by step

Step 1. Open Cowork in your desktop app, pointed at your Switched-On folder. Opus for this one.

Step 2. Paste this brief:

Create a file at the top of this folder called CLAUDE.md. The purpose of CLAUDE.md is to be the master brief that you (Claude) read at the start of every session opened from this folder.

The file should include:

1. Who I am, in 3-5 lines (you can ask me). 2. What this folder is for (Switched-On, my personal AI workspace). 3. How to behave in this folder. Specifically: use Australian / British English spelling. No em dashes. No corporate jargon. Be direct. Push back when I’m wrong. Ask clarifying questions before generating anything substantial. Save outputs to outputs/ using YYYY-MM-DD_description.md naming. 4. Where to look for things: outputs/, skills/, reference/. 5. A note that MEMORY.md is the running log of what we’ve learned together, and you should read it too at the start of every session.

Ask me the questions you need to write section 1 (about me). Then draft the whole thing. Show me before saving. I’ll edit.

Step 3. Answer Claude’s questions about you. Don’t overthink. Sentences not paragraphs. Use Wispr Flow if it’s faster to talk than type.

Step 4. Read the draft. Edit anything that doesn’t sound like you. Approve.

Step 5. Now MEMORY.md. Paste this:

Now create a file at the top of this folder called MEMORY.md. The purpose of MEMORY.md is to be a running log of what we’ve learned together across sessions. Pattern preferences. Corrections I’ve made. Things I’ve told you to remember. Decisions I’ve made that I want you to remember next time.

Start with three example entries based on what we’ve worked on across the past two weeks. Use this format: - YYYY-MM-DD: [what was learned, in one sentence]

Going forward, you’ll add to this file whenever I tell you to “remember that” or when you notice a pattern worth recording.

Step 6. Read the three starter entries. Add anything obvious you’d want to remember from the past two weeks. Save.

Step 7. Set the global instruction. Open Claude desktop settings. Find Custom Instructions (or System Prompt or Behaviour, depending on which version you’re on).

Paste this paragraph:

When opening any folder containing a CLAUDE.md file, read CLAUDE.md and MEMORY.md first before doing anything else. Treat their contents as instructions for how to behave in that folder. Confirm at the start of the session that you’ve read them.

Step 8. Test it. Close Cowork. Reopen. Open your Switched-On folder. Type:

What do you know about me and how I work? Summarise what you’ve read in CLAUDE.md and MEMORY.md.

Claude should come back with a clean summary. If it does, the spine is in. If it doesn’t, the section below covers what to check.

If Step 8 didn’t work

Four things to try, in order:

1. The filename. Make sure it’s CLAUDE.md and not CLAUDE.md.txt. Mac TextEdit sometimes adds the .txt extension silently. In Finder, the .md extension should be visible at the end of the filename.

2. Close and reopen Cowork. The global instruction only takes effect on new sessions started after you saved the setting.

3. The settings menu might be called something else. Different Cowork versions call it Custom Instructions, Personal Settings, or Behaviour. Try all three.

4. Curly quotes vs straight quotes. If the pasted instruction in your settings menu shows curly quotes rather than straight ones, retype the quote characters directly.

If none of the four work, hit reply and tell me what’s happening. The fix is almost always one of these.

Two weeks in. How’s it landing?

Quick favour before the weekend.

This is the end of Week 2. Two weeks ago, AI was a browser tab and a chat window. Now you’ve installed Cowork, connected two services, built a dashboard, killed a few Sunday Scaries, read a document in five minutes, decided something you’d been weighing, and given Claude its first proper brief on who you are.

That’s not nothing.

I want to hear how it’s going. Three questions, if you’ve got a minute over the weekend:

1. What’s the moment from these two weeks you’d quote to a friend? The bit you’ve already told someone about.

2. What’s the thing you haven’t quite cracked? Where the friction is.

3. What’s one tool, habit or file you’d like to see covered in Week 3 or 4? I’m shaping the back half this weekend.

Hit reply. DM me on Instagram. Comment on the post. They all come to me. I read every one.

One last thing. The women getting most from this challenge are the ones who’ve told a friend. The first time you explain it to someone else is usually when it lands properly. If there’s a woman in your life who’d want this: nishma.co/30daychallenge.

A safety note

CLAUDE.md and MEMORY.md sit at the top of your folder and Claude reads them every session. Treat them like the brain you actually want Claude to have. Don’t write anything in them you wouldn’t be comfortable Claude using as context in any future chat.

Same backup discipline as yesterday. iCloud, Dropbox, whatever you use. The spine is now part of your professional setup.

What you’ll notice

The next chat in this folder feels different.

You don’t have to re-explain who you are. You don’t have to specify “use British English” or “save to outputs/” or “no em dashes.” Claude already knows. The first session after the spine is in is a small wow moment that compounds across every session after.

The other thing you’ll notice (probably tomorrow): you start using Claude for more things, because the cost of starting a new task has dropped. You’re not warming the model up. The model already knows.

Why this works

Most AI tools fail you because every session starts from zero. You re-explain who you are. You re-state your standards. You re-paste your style. Half of the prompt is set-up.

CLAUDE.md plus MEMORY.md plus a global instruction means the set-up is permanent. You write it once. It applies forever. Every prompt from here on is shorter and sharper, because the context is already loaded.

This is the gap between people who get a lot out of AI and people who don’t. It isn’t intelligence. It isn’t tool choice. It’s whether the AI knows you when you arrive. From today, yours does.

Key takeaway

Two text files at the top of your Switched-On folder. One paragraph in settings. From tomorrow, Claude opens this folder and instantly knows who you are, how to behave, and what you’ve learned together.

Twenty minutes today. Compounds across every session for the rest of your working life.

Where this is heading

Week 2 closes here. Week 3 begins tomorrow.

The first two weeks were about giving Claude work to do. Week 3 is about telling Claude who YOU are, properly. Six brain files across the week, each one a different facet of you and your business. Each one saved to the spine. Each one read automatically forever after.

It starts tomorrow with the most important one.

Tomorrow on Switched On

Day 15: Tell AI who you are, properly.

Saturday edition. Ten minutes of Claude interviewing you. Walk out with a one-pager that briefs any AI like a new hire. The first brain file. The one everything else hangs on.

Nishma x

P.S. How did today land? Three buttons, thirty seconds → ​Tap one​

I read every response. The data shapes the rest of the challenge.

Catching up? Days 0 to 13 are on the Switched On Substack archive ​HERE​.

Got a friend or a colleague who this might be useful for? Send her to nishma.co/30daychallenge

SwitchedOn #SwitchedOnAIChallenge

Welcome to Day 13 of the challenge.

Today’s job is small. Three folders. Eight minutes.

It will look like nothing. By next week it’ll be the foundation that makes everything you build with Claude faster, sharper and reusable.

This is the day the word “infrastructure” goes in. Don’t be put off by it. Infrastructure is not technical. Infrastructure is naming files in folders so Claude knows where to put things and where to find them. That’s the whole magic.

A word before we start

So far your Switched-On folder has whatever you’ve saved into it. The Day 5 subscriptions table. The Day 7 dashboard. The Day 8 Saturday sort. The Day 9 meeting briefs. The Day 10 daily plans. The Day 11 doc summary. The Day 12 decision file.

It’s already a lot. By Day 30 there’ll be a lot more.

Without structure, “the Switched-On folder” becomes a mess you can’t find anything in. With structure, you have a working brain.

Why this folder works in any AI tool

The reason we’re using plain text files (.md) and standard filenames is not because Claude is special. It’s because the whole setup is portable.

CLAUDE.md is named for Claude, but a folder structured like this works with any AI tool that reads files. Drop the same Switched-On folder into ChatGPT, Gemini, Cursor, Perplexity Spaces, or whatever’s next, and a sensible AI will read the context. Your brain about you isn’t locked to one product. It’s yours.

Same for the naming convention. Lowercase, hyphens, dates first. Works in any operating system, any cloud service (iCloud / Dropbox / OneDrive / Google Drive), any text editor. No proprietary format. No lock-in.

You’re building a personal AI infrastructure that survives any tool change. That’s the real durability play.

The structure

Three sub-folders is the entire system:

• outputs/ for everything Claude produces for you (the daily plans, the doc summaries, the decisions, the meeting briefs, the Saturday sorts)

• skills/ for the agents you’ll build in Week 4. A ‘skill’ is a saved set of instructions Claude can run on demand, like a recipe. The skills/ folder holds them. You’ll build your first one on Day 26.

• reference/ for documents you drop in for Claude to read (your contracts, your audience research, your brand guidelines, anything you want Claude to be able to refer back to)

Three folders. That’s it. Today we make them and tell Claude where things go from now on.

Step by step

Step 1. Open Cowork in your desktop app, pointed at your Switched-On folder.

Step 2. Move the files you’ve made so far into outputs/. Don’t do this yourself. Ask Claude:

Look at the files currently in this folder. Move everything I’ve saved from Days 5 through 12 (the subscriptions table, the dashboard HTML, the Saturday sort, the meeting briefs, the daily plans, the doc summary, the decision file) into a new sub-folder called outputs/. Don’t move CLAUDE.md or MEMORY.md if they exist (we make those tomorrow).

Watch what Claude does. If something gets moved you’d rather leave at the top level, just tell it to move it back.

Step 3. Create the other two sub-folders. Ask Claude:

Now create two more sub-folders alongside outputs/: skills/ and reference/. Leave them empty for now. We use skills/ in Week 4 and reference/ later this week.

Step 4. Tell Claude the rule for the future. Paste this:

From now on, when you produce anything for me, save it to outputs/ using the naming convention YYYY-MM-DD_short-description.md. Use lowercase, hyphens not spaces, today’s date. So a daily plan today would be 2026-XX-XX_daily-plan.md. Confirm you’ve understood.

Step 5. Drop something into reference/.

Anything you’d want Claude to be able to refer back to in future chats. Examples:

• Your brand guidelines doc (if you have one)

• A past LinkedIn post you’re proud of

• A long-form bio

• A pricing sheet

• An audience research document

• A document you keep being asked to reference

Pick one. Drag it into reference/. (You can add more over time.)

Step 6. Test the structure. Start a new chat and ask:

What’s in my Switched-On folder right now? List the structure.

Claude reads the folder, tells you the structure. If everything’s where it should be, you’re done.

A safety note

This is admin, not connection. No new permissions today. Claude is only doing things inside the folder you’ve already given it access to.

Two small good habits to install today:

1. Don’t put anything in this folder you wouldn’t be comfortable having Claude read. This is the working space, not the secure archive. If a document is genuinely sensitive, keep it elsewhere.

2. Back the folder up. iCloud, Dropbox, OneDrive, whatever you use. Switched-On is becoming the brain. Treat it like one.

What you’ll notice

The folder feels cleaner than it did yesterday. That’s it. That’s the entire visible payoff today.

The actual payoff isn’t visible until next week, when Claude starts opening this folder and immediately knowing where things go and where to look. From then on, every prompt is faster. Every output is findable. Every reference is reusable.

You don’t notice infrastructure until it’s not there.

Why this works

Naming is structure. Structure is memory. Memory is the difference between Claude as a chat tool and Claude as a colleague.

Most AI tools fail you because everything lives in chat history that nobody can find again. Cowork plus a folder structure flips that. Outputs are files. Files have names. Names are findable. Tomorrow’s Claude opens this folder and inherits everything today’s Claude built.

This is the move 95% of women using AI never make. The 5% who do make it are the ones you keep meeting at events and wondering “how do they get so much out of this.” It’s this.

Key takeaway

Three sub-folders. Outputs, skills, reference. One rule: save everything to outputs/ with a dated filename. Eight minutes.

Looks like nothing. Is everything.

Where this is heading

Tomorrow is the day this structure earns its name. We write the two text files that turn the folder from “place where files live” into “place that briefs Claude on you every time it opens.”

That’s the spine.

Tomorrow on Switched On

Day 14: Write the two files that make AI smarter forever.

Twenty minutes. Two text files. One paragraph in settings. From tomorrow on, Claude opens this folder and instantly knows who you are.

Nishma x

P.S. How did today land? Three buttons, thirty seconds → ​Tap one​

I read every response. The data shapes the rest of the challenge.

Catching up? Days 0 to 12 are on the Switched On Substack archive ​HERE​.

Got a friend or a colleague who this might be useful for? Send her to nishma.co/30daychallenge

SwitchedOn #SwitchedOnAIChallenge

Welcome to Day 12 of the challenge.

There’s a decision you’ve been carrying. A software tool you’re choosing between. A hire you’re weighing. A vendor. A venue. A format. Maybe an offer on the table you keep meaning to say yes or no to.

It’s been a week longer than it should have been.

Today we put it to bed.

A word before we start

Most “I’m weighing this” is not really weighing. It’s circling. The reason it’s been a week is not that the decision is hard. It’s that you haven’t been forced to be specific about your criteria yet.

The cure isn’t more thinking. It’s a sharper question.

Today Claude plays the role of the brilliant strategist who, instead of trying to recommend anything, makes you answer five questions you’ve been avoiding. Once you’ve answered them, the decision usually makes itself.

This is the version of AI as a thinking partner most people miss. It’s not “AI tells me what to do.” It’s “AI asks me what I should have asked myself.”

Step by step

Step 1. Pick the decision. Real one. Currently on your plate. Has been there longer than a week.

Don’t pick the biggest decision of the year. Pick one that’s mid-stakes and overdue. Software tool, hire, vendor, format, venue, offer to say yes or no to. The kind of thing that’s been sitting on the back of your brain.

Step 2. Open Cowork. New chat. Paste this brief:

I’m choosing between [Option A] and [Option B] for [purpose].

Context: [3-4 lines on what’s at stake, what’s making it sit, what I’ve already half-decided.]

Don’t recommend yet. Step one: ask me FIVE questions that would help you make a sharper recommendation. The kind a real strategist would ask, not the obvious ones I’ve already asked myself. One question at a time. Wait for my answer before moving to the next.

Hold the recommendation until after question five.

Step 3. Answer the five questions. One at a time. Don’t rush. The questions are the work.

A few will feel uncomfortable. That’s the signal you’re answering the right ones. Use ​Wispr Flow​ if you want to talk through your answers instead of typing them.

Step 4. After question five, paste this:

Now compare the two options against the criteria my answers implied. Score each on a 1-10 scale per criterion. Tell me which to pick. If it’s close, tell me what would tip it the other way.

Be direct. Don’t hedge.

Step 5. Read the recommendation.

You’ll either feel relief (the answer matches what you were already leaning toward, you just hadn’t said it out loud) or surprise (the answer is the opposite and you need to sit with it).

Both are useful. Don’t override the surprise on instinct. Sit with it for an hour. If at the end of the hour you still disagree, override. Tell Claude why. Save that note. The reasoning matters more than the choice.

Step 6. Save the decision.

Save this whole conversation as 2026-XX-XX-decision-[short name].md in my Cowork folder.

This becomes the receipt. Six months from now, when you’ve forgotten why you chose the way you did, the file remembers.

Step 6.5. Draft the message.

The decision is the inside work. The message is the outside work. Day 12 isn’t done until the outside work lands. Ask Claude:

Draft the message I’ll send to [the person waiting] explaining the decision and what happens next. Three to five lines. My voice. No hedging.

Read it. Edit if needed. Step 7. Send it.

Don’t end the day with the message drafted but un-sent. Email, Slack, text, whatever the channel. Close the loop today.

Step 8. Add to MEMORY.md.

Before you close, add one line to MEMORY.md (or jot it down to add tomorrow if you haven’t built MEMORY.md yet):

Decision pattern that works for me: five questions before recommendation, then the message drafted immediately.

Reinforces the pattern for next time.

A safety note

If the decision involves people (a hire, a redundancy, a contract you’re walking away from), Claude is excellent on the criteria and clumsy on the humanity. Use it for the criteria step. Don’t use the recommendation as your script when you go to talk to anyone.

Same for anything legal, medical or financial with real consequences. Claude is a clarifier of your thinking, not a substitute for advice from the relevant professional.

What you’ll notice

You’ll already know the answer after question three or four.

That’s normal. The questions surface what you’d been holding back from yourself. By the time Claude recommends, the recommendation is a formality.

The other thing you’ll notice: the relief of having decided is bigger than the decision deserved. That’s the open-loop tax you’d been paying. Every day the decision sat on the back of your brain, it was costing you energy you weren’t accounting for. The five-minute exercise reclaims it.

Run this once a month on whatever’s been sitting longest. Compound effect across a year is real.

Why this works

Decisions stall because the criteria aren’t named. Once the criteria are named, most decisions are obvious. A good strategist’s job is to surface the criteria, not to recommend.

Claude is good at this because it has no skin in the game and no relationship to protect. It just asks the question your inner circle has been too polite to ask.

The pattern is the same as the other days. Claude does the assembly (in this case: asking the right questions). You make the call. The call is faster because the assembly is done.

Key takeaway

Five questions before the recommendation. The decision you’ve been carrying, decided, in twelve minutes. The receipt saved for future-you.

This is what AI as a thinking partner actually means.

Where this is heading

You’ve used AI on real tasks all week. Tomorrow we set up the structure that holds everything you build with it. Three folders. Eight minutes. The foundation Week 2 closes with.

Tomorrow on Switched On

Day 13: Set up the folder that holds your AI brain.

Three empty folders. That’s the infrastructure. It looks like nothing today. By the end of next week it’ll hold the brain about you.

Nishma x

P.S. How did today land? Three buttons, thirty seconds → ​Tap one​

I read every response. The data shapes the rest of the challenge.

Catching up? Days 0 to 11 are on the Switched On Substack archive ​HERE​.

Got a friend or a colleague who this might be useful for? Send her to nishma.co/30daychallenge

SwitchedOn #SwitchedOnAIChallenge

Welcome to Day 11 of the challenge.

There’s a document on your desk. Maybe a board pack. Maybe a contract. Maybe a 47-slide pitch deck from a vendor. Maybe a long report you’ve been told to read.

You haven’t opened it. It’s been sitting there since last Wednesday.

Today we close that loop. Not by you finally sitting down with it. By Claude reading it for you and telling you only what you actually need.

Five minutes. Decision made. Move on.

A word before we start

Most AI “summarise this” prompts produce a summary. A summary is not what you want.

A summary tells you what’s IN the document. It doesn’t tell you what to DO. Half the time it doesn’t even tell you what matters.

You’re not a librarian. You’re a decision-maker reading the document because you need to act, push back, sign, decline, approve, escalate, walk into a meeting prepped. The brief has to ask Claude for that. Sharply.

Today’s prompt is the most powerful one in the challenge so far. Memorise the shape of it. Use it on every long document for the rest of your working life.

Step by step

Step 1. Pick a real document you’ve been avoiding. Not a hypothetical one. The actual one on your desk or in your downloads folder. Board pack, contract, deck, report. Anything ten or more pages you’d rather not read.

Step 2. Open Cowork in your desktop app. Opus for this one. (The thinking is heavier today.)

Step 3. Drag the file into a new chat. Or use the + button to attach. PDF, Word, deck, plain text, all work. If it’s a Google Doc, export to PDF first.

Step 4. Paste this brief:

I need to know what’s in this document but I don’t have 45 minutes. Tell me, in this order:

1. The one thing if I only read one sentence. 2. The three things if I read for two minutes. 3. What to question or push back on. Anything that doesn’t add up, contradicts itself, or feels weak. 4. Anything in here that contradicts something earlier in the doc. 5. The questions I should ask before I respond / approve / sign / decide.

Be ruthless. If a section is filler, say so. If a claim is unsupported, say so.

Context for you: [in 2-3 sentences, tell Claude WHY you’re reading this. Who’s asked for your response. What decision you’re being pulled into. What you suspect already.]

Before you start, ask me three clarifying questions about the context.

Step 5. Answer Claude’s three questions. (This step matters. The questions force Claude to pin down what you’re really looking for. Without them, you get a generic read.)

Step 6. Read the output.

You’ll notice the structure pulls you forward. The one-sentence read first. The three-minute read second. The pushback third. By the time you’ve finished reading Claude’s brief, you’ve already formed an opinion.

Step 7. For each item flagged as “question or push back on,” ask Claude:

Quote me the exact line from the document where this happens. Give me the page number or section if there is one.

That’s the receipt you’ll use in the meeting, the reply, the escalation.

Step 8. Save the brief.

Save this as 2026-XX-XX-doc-[short name].md in my Cowork folder.

When the meeting happens or the email gets sent, you’ve got the read. When someone challenges you, you’ve got the receipts.

A safety note

If the document is genuinely confidential (legal, M&A, board-sensitive, employee data), think before you drop it into a chat. Two checks:

1. Is it covered by your company’s AI policy? Most enterprise Claude accounts are now approved for confidential material. Personal Claude Pro is treated as your own device. If you’re not sure, ask IT or check the policy doc.

2. Would you read it on a train with someone sitting next to you? If yes, drop it in. If no, redact the sensitive sections before you attach.

When you’re done, delete the chat. The document goes with it.

What you’ll notice

The summary is shorter than you’d write. By a lot.

A 60-page board pack collapses to four or five lines that matter. The rest is genuinely filler, repetition or framing. Once you see this happen once, you trust it. Twice, you stop reading long docs the slow way.

The pushback is the unlock. The “things that don’t add up” list will surface contradictions you’d have missed because you’d have read in a straight line. Claude reads non-linearly. It cross-checks. It catches the buried number that doesn’t match the earlier slide.

You walk into the meeting having read for five minutes and you’re sharper than the person who read for forty.

Why this works

The friction with long documents was never the reading. It was the reading PLUS the formulating-an-opinion PLUS the remembering-what-mattered PLUS the cross-checking. Four jobs disguised as one.

Claude does the assembly. You make the call. The call is faster because the assembly is done. Same pattern as Day 5 (statement reading), Day 7 (inbox cross-check), Day 8 (inbox triage), Day 10 (planning). You’ll see it everywhere from here on.

Side note: when NotebookLM beats Claude for this

If the document is something you need to LEARN (a complex regulation, an academic paper, a deeply technical brief from a discipline you don’t know), Google’s NotebookLM is genuinely better than Claude for the learning task. It builds a chat interface trained on the document so you can keep asking questions, and produces study-guide-style summaries and audio overviews.

For DECISION-PREP, which is what today is, Claude is sharper.

Different jobs, different tools. Knowing which to reach for is the next layer of fluency.

Key takeaway

The “what to DO” prompt, not the “summarise” prompt. Five minutes. The doc you’ve been avoiding, read, with pushback and questions ready.

Use this prompt structure on every long document for the rest of your working life.

Where this is heading

You’ve made AI do the reading. Tomorrow we make it do the deciding. Or rather, we use it to make the decision you’ve been weighing for a week, the right way.

Tomorrow on Switched On

Day 12: Decide the thing you’ve been weighing.

Five questions. Five minutes. The A-or-B decision sitting on your plate, off your plate, today.

Nishma x

P.S. How did today land? Three buttons, thirty seconds → ​Tap one​

I read every response. The data shapes the rest of the challenge.

Catching up? Days 0 to 10 are on the Switched On Substack archive ​HERE​.

Got a friend or a colleague who this might be useful for? Send her to nishma.co/30daychallenge

SwitchedOn #SwitchedOnAIChallenge

​

Welcome to Day 9 of the challenge.

The Sunday you spent NOT thinking about Monday. And then it’s 7pm and Monday is in your face anyway.

You know the meetings. You’ve seen the calendar at a glance. You can’t quite remember who’s in the room for the 10am, you’ve half-thought about what you’ll say in the 2pm, and there’s one meeting you genuinely don’t know what it’s about yet.

Eight minutes today and all three feel sorted.

A word before we start

Today is your second connector. Same flow as Day 7, same permissions stack, same read-only-first principle. Once you’ve connected one calendar to Claude, you’ll never want to walk into a meeting cold again.

This is the day people quote back to me most when they describe their first weeks with Cowork. The “I felt like I’d had a chief of staff overnight” moment is usually this one.

A note for those of you on Outlook or Apple Calendar: the connector works the same way. The clicks below are Google-specific, but find the equivalent for your provider and the brief is identical.

Step by step

Step 1. Open Cowork in your desktop app. Settings. Connectors.

Step 2. Find Google Calendar (or Outlook / Apple Calendar). Sign in.

​

​

​

Step 3. Grant READ-ONLY access. Same rule as Gmail. Claude needs to read your calendar, nothing more. If the screen offers a choice, untick anything that says create, modify or delete events. Read-only is enough for everything in this brief.

​

Step 4. Pick the three meetings from next week that matter most. Not all of them. The three you’d actually want to walk in sharper to.

If you’re not sure which three: any meeting where the room contains someone senior to you, any meeting where money or scope is on the table, any meeting where you’re presenting, any meeting you’ve already half-prepped for. Pick three. Don’t optimise.

​

Step 5. In a new Cowork chat, paste this brief:

Read my Google Calendar for next week.

Focus on these three meetings: [paste meeting titles and dates, or describe them in your own words].

​

For each meeting, give me a one-page prep brief. Specifically:

​

1. Who’s in the room. Names, roles, anything I should know from prior emails (use Gmail to cross-check).
2. What this meeting is for. The invite text, the actual goal, any context from the last conversation.
3. Last conversation context. Search Gmail for the most recent thread with this group. Two sentences on where things were left.
4. What’s at stake. What I should be ready for. Decisions on the table, sensitivities to watch for.
5. Three opening questions. Sharp ones I could lead with. The kind that change the temperature of the room.

​
Before you start, ask me three clarifying questions about what’s missing from the calendar invites themselves.

Step 6. Answer Claude’s three questions. Let it run.

(If this is the first brief you’ve given Claude that touches both your calendar AND your inbox, you’ll see it stitch the two together. That’s the unlock. Your work doesn’t sit in one app, and your colleague no longer pretends it does.)

Step 7. Read the prep brief for each meeting.

For each one, do three things:

• Tell Claude what it missed. Context only you have. The actual politics of the room. The fact that the senior person hates being interrupted. The fact that you’ve been chasing this for six weeks.

• Cut the questions you wouldn’t actually ask. Claude’s three opening questions will usually need one swap. Trust your instinct.

• Add the one thing you want to land. What’s the sentence you want the room to remember from you?

Step 8. Save the briefs.

Ask Claude:

Save each meeting brief as a separate file in my Cowork folder. Name them 2026-XX-XX-prep-[meeting name].md. Put the date of the meeting at the top of each one.

You now have three files you can open on Monday morning, five minutes before each meeting. Re-read, pick the three things you’ll bring in, walk in.

A safety note

Calendar contents are sometimes more sensitive than you’d think. Your week tells the story of who you’re meeting, when, where, and sometimes why. Claude has read-only access to that data. It cannot create, change or share anything.

If a specific meeting is private (a board sub-committee, a difficult one-to-one, anything legal), exclude it from the brief by name: “Skip anything titled [meeting name].” Claude won’t touch it.

Revoke calendar access the same way you’d revoke Gmail. Settings, Connectors, Disconnect. Or in Google: myaccount.google.com, Security, Third-party apps with account access, find Claude, Remove access. Both work.

What you’ll notice

The prep for the meeting you’ve half-thought-about feels disproportionately good. The cross-check between calendar and inbox surfaces context you’d otherwise have walked in without.

The third meeting (the one you genuinely didn’t know the agenda for) usually becomes the most useful brief. Claude pulls the invite text, scans the prior thread, and tells you what it’s about before you have to ask anyone.

The single biggest shift: you stop walking into meetings as the person catching up and start walking in as the person already two steps ahead. Same calendar. Different posture.

Why this works

Most of meeting prep is information assembly. Who’s there, what was last said, what’s at stake. The thinking bit, your angle, your push, your read on the room, is yours. The assembly bit, which is what makes prep feel heavy, isn’t. That’s Claude’s.

Once you’ve separated the two, eight minutes is enough. The brief reads as if you’ve spent half an hour, because the assembly LOOKS like the work. It isn’t.

Key takeaway

One connector. One brief. Three meetings, briefed. Sunday afternoon. Monday morning sorted.

You now have a routine you can run every Sunday evening. Calendar plus inbox plus a one-page brief per meeting. Eight minutes. The version of yourself that walks into next week is the prepared one.

Where this is heading

Two connectors in. Two days of “wow.”
​You’ve crossed the line from “AI is a chat tool” to “AI is a colleague who lives in my work.”

Tomorrow we tie everything you’ve done so far into a daily routine.

Tomorrow on Switched On

Day 10: The five-minute end-of-day routine.

Sunday Scaries don’t only happen on Sundays. They happen at 11pm on a Tuesday too. Tomorrow we install the daily habit that ends them for good. Five minutes before you close the laptop. Tomorrow morning, a plan instead of a panic.

Nishma x

​

P.S. How did today land? Choose one answer (10 seconds max) → Tap one​

I read every response. The data shapes the rest of the challenge.

​
​

Catching up? Days 0 to 8 are on the Switched On Substack archive HERE.

​

Got a friend or a colleague who this might be useful for?
Send her to nishma.co/30daychallenge

SwitchedOn #SwitchedOnAIChallenge

Welcome to Day 10 of the challenge.

Yesterday you prepped Sunday for Monday. One-off. A specific moment for a specific knot.

Today is the version you do every day for the rest of your working life.

Five minutes at the end of every workday, before you close the laptop, that turns tomorrow morning’s panic into tomorrow morning’s plan. Same calendar and Gmail connectors you set up over the weekend. New brief. New habit.

The Sunday Scaries are real on a Sunday. They’re also real at 11pm on a Tuesday. The cure is the same.

A word before we start

Most planning advice tells you to do it in the morning. I don’t agree.

Mornings have momentum. They have energy. They have a window when you should be doing the most important work, not deciding what the most important work is. By the time you’ve finished deciding, the morning is gone.

Tomorrow’s plan belongs to tonight. Five minutes at the end of today’s work, while context is still warm, while you can still remember what’s actually in the calendar and what’s actually overdue. Future-you reads it tomorrow morning, decides nothing, just does.

This is the rhythm Week 2 was building toward. Connectors plus a prompt plus a saved file. Five minutes. Compounds across a year into something that genuinely changes how you work.

Step by step

Step 1. Pick your trigger. 5pm, 6pm, 7pm. Whenever your work day actually ends. The same time every day matters more than the specific hour.

Step 2. Open Cowork in your desktop app.

Step 3. Start a new chat. Paste this brief:

Build me tomorrow’s plan. Read tomorrow’s calendar and today’s unactioned inbox.

Give me, in this order:

1. Three things that MUST move forward tomorrow. Specific, not generic. From the calendar, the inbox or anything I’ve flagged. 2. Two things I can do tomorrow but don’t have to. Bonus territory. 3. One thing I should say no to or push back on. Be direct. Tell me what’s not earning its place in tomorrow’s day. 4. The single most important meeting tomorrow, and the one thing I should pre-think before walking in.Anything in tomorrow’s calendar that shouldn’t be there. 5. Anything I’d want to reschedule or decline if I were honest.

Before you start, ask me three clarifying questions.

Step 4. Answer Claude’s three questions. Let it run.

Step 5. Read the plan.

Push back where it got it wrong. Sample correction:

The meeting with [name] is more important than you think. Move it up. The [other meeting] isn’t actually mine to lead. Take it off the priority list.

Step 6. Save the plan. Ask Claude:

Save this as 2026-XX-XX-plan.md in my Cowork folder. One file per day. I’ll open it first thing tomorrow.

Step 7. Close the laptop. Evening starts now.

Tomorrow morning

Before email. Before Slack. Before anything.

Open the file. Read for two minutes. Decide which of the three priorities you’ll start with. Walk into the day.

You’ll find it odd at first. The instinct is to open the inbox. Resist. The plan you made last night already accounts for what’s there. Trust it.

A safety note

No new connectors today. You’re using yesterday’s calendar connector plus Day 7’s Gmail connector. Both still on read-only. Both still revocable from Settings.

If you’d rather Claude only see specific labels in Gmail when building tomorrow’s plan (work labels only, say), add this to your brief: “Only read emails labelled Work / Clients / Urgent. Skip anything in personal labels.”

What you’ll notice

The “three things that must move” list is shorter than you think it should be.

That’s not a bug. That’s the shift. Most days have three real priorities and a lot of noise. The five-minute brief surfaces the three. The noise is still there, but it stops being the thing that defines your day.

The second thing you’ll notice: evenings get quieter. The mental open loop of “I should probably think about tomorrow” closes when you actually do. The five minutes buys you the rest of the night.

The third thing: tomorrow morning starts faster. No 20-minute warm-up while you remember what you were supposed to be doing. The file already knows.

Why this works

Planning and doing are different jobs. We try to do both at once (in the morning, badly) and resent both.

Splitting them is the move. Last-night-you plans. This-morning-you does. The handoff is a saved file. Both versions of you are working at the time of day each is sharpest.

It’s the same Cowork pattern again. Claude assembles. You decide. The decision is faster because the assembly is done.

Going further (optional, +10 mins)

Once you’ve nailed the base version, layer in any of these:

1..Draft the follow-up emails. After the plan, ask:

Draft the three follow-up emails I should send tonight to close loops from today’s meetings.”

Saves you 20 minutes tomorrow.

Saves you twenty minutes tomorrow morning.

2. Cross-check Granola, Fyxer or Otter. If you use a note-taker or meeting-notes AI, drop in today’s notes alongside the brief:

Add anything from these notes that I committed to and haven’t actioned. Surface it in tomorrow’s priorities if it’s overdue.

3. Surface the buried commitments. Once a week, ask:

“Read the last seven days of sent mail. List every time I said ‘I’ll send you’ or ‘I’ll follow up’ and never did.”

This is the one people gasp at. The commitments you forgot you made, surfaced and either actioned or apologised for. Run weekly.

Key takeaway

Five minutes at the end of every workday. Calendar + inbox + one brief. Tomorrow’s plan as a saved file. Open it in the morning before anything else.

Mornings are won the night before.

Where this is heading

Two connectors. One daily ritual. You’ve now got the rhythm Week 2 was about. The rest of the week is the high-value tasks the rhythm makes possible.

Tomorrow on Switched On

Day 11: Read a 60-page research document or board pack in five minutes.

The boring document on your desk doesn’t need 45 minutes. It needs a sharper prompt. Tomorrow we hand off the doc you’ve been avoiding and Claude tells you only what you need to act on.

Nishma x

P.S. How did today land? Three buttons, thirty seconds → ​Tap one​

I read every response. The data shapes the rest of the challenge.

Catching up? Days 0 to 9 are on the Switched On Substack archive ​HERE​.

Got a friend or a colleague who this might be useful for? Send her to nishma.co/30daychallenge

SwitchedOn #SwitchedOnAIChallenge

​

Welcome to Day 8 of the challenge.

It’s Saturday. Different rules apply.

The Monday-to-Friday version of this challenge is about giving AI work to do. Today is about clearing the personal admin pile you’ve been pretending isn’t there.

The kettle. The mug. The laptop reluctantly open. The inbox that hasn’t seen a proper sort since (insert your own personal worst-case month here).

Twenty minutes. Then back to the weekend.

A word before we start

Yesterday you connected Claude to Gmail for one specific job: catching the annual subs the bank statement missed. Today you point the same connector at a different problem. The inbox itself.

This is the move people skip and then regret. Most of us treat the inbox as a to-do list and then drown in it. Claude doesn’t see it that way. Claude sees a triage problem. Who wants what, when, and whether it actually needs you.

Same connector. Same permissions. Different intent. That’s the lesson hiding underneath today’s task: one connector unlocks dozens of briefs once you stop seeing it as “the Gmail one” and start seeing it as “the inbox-reading one.”

A note for those of you with separated work and personal accounts: today is the personal one. Save the work inbox for Monday. We fold the work version into the end-of-day routine on Day 10. If you’ve only got one Gmail account that does both, fine. Run it on the lot.

Step by step

Step 1. Make a tea. (Saturday rules.)

Step 2. Open Cowork in your desktop app. Opus if it’s available to you. (Day 4’s choice.)

Step 3. Start a new chat. Paste this brief:

Read my Gmail inbox from the past 14 days. Don’t reply or action anything. Triage every email into four buckets, in this order:
​

1. Needs me this week. Real asks, real decisions, real people waiting for an actual response.
2. Already handled. Receipts, booking confirms, “you have a dentist appointment” reminders. Noise I just need to know is there.
3. Saturday sort. Personal admin I should action this weekend while I’m here. School forms, contract signatures, online returns, the booking I keep meaning to make.
4. Bin or unsubscribe. Newsletters I don’t read. Promotional emails I never opened. Notifications I never asked for.
​

For each email tell me: who it’s from, what they want, which bucket, why. Be specific. Don’t generalise.
​

At the end: how many emails in each bucket, the one most-overdue thing in bucket 1, the most useful thing I can unsubscribe from in bucket 4.
​

Before you start, ask me three clarifying questions about my inbox. Then run.

(The “ask three questions first” line is Day 3’s move. Use it on every brief from here on. It saves you the wasted output of an AI that guessed wrong.)

Step 4. Answer Claude’s three questions. Let it run.

Step 5. Read the table.

For each row in Needs me this week, decide:

• Action now (less than 90 seconds): just do it

• Action this week (more than 90 seconds): leave it, with a flag

For each row in Saturday sort:

• Pick three you’ll actually do today

• Leave the rest for next weekend

For each row in Bin or unsubscribe, ask Claude:

Give me a one-line unsubscribe brief for [sender]. Where do I click?

Step 6. Push back where Claude got it wrong.

Claude doesn’t know everyone in your inbox. It will sometimes put a real friend’s note into “noise” or a routine confirmation into “needs me.” Tell it. Sample correction:

Wrong on [sender]. That’s [my sister / my accountant / the school]. They go in [bucket]. Anything else from them defaults to that bucket too.

This is the most important step of the day. Every push-back is a piece of training. Claude gets sharper on your inbox the more you do it.

Step 7. Save the table back to your Cowork folder.
Name it 2026-XX-XX-saturday-sort.md (your date).

We come back to this on Day 10 when we turn the inbox sweep into an end-of-day routine. The version you build today is the standalone personal sort. The work version folds in then.

A safety note

The connector is the same one you read about yesterday. Read-only access only. Claude can read what’s in your inbox. It cannot send, delete or modify anything. The bucket sort is descriptive only. Nothing in your inbox changes until you decide to change it.

If anything in your inbox is genuinely sensitive (legal correspondence, anything medical, anything financial you’d rather Claude didn’t see), exclude it by label: “Skip anything labelled Legal or Medical or Private.” Claude won’t touch it.

Label-level scoping (only granting access to specific Gmail folders rather than the whole inbox) is on the cards for Week 3, when we go deeper on connector hygiene. For today, the read-only flag plus exclusions by label is the line of defence.

What you’ll notice

Bucket 4 is usually the biggest. By a long way.

Most personal inboxes are 80% noise. The noise is what makes “I’ll just go through my emails” feel like a four-hour job. Claude pulls the noise out and you’re left with the 20% that’s actually you.

The second thing you’ll notice: bucket 1 is smaller than you thought. The mental image of “my inbox is unmanageable” usually isn’t matched by the data. It just FEELS unmanageable because it’s all mixed in.

The third thing: at least one item in bucket 3 has been overdue for months. Saturday morning is the day to action it.

Why this works

Triage isn’t reading. Reading is reading.

Most women lose hours to inbox stress because we’re trying to do both at once. Read everything AND decide what to do with it. Claude separates the two. Claude reads. You decide.

That’s the Cowork pattern again. Same as Day 5 (statement reading). Same as Day 7 (subscription cross-checking). The friction isn’t the decision. The friction is the line-by-line groundwork before you can decide. Claude eats that.

Key takeaway

The Saturday sort. Same Gmail connector, different brief. Four buckets, one pass, twenty minutes.

Most of your inbox is noise. Claude tells you which 20% isn’t.

Where this is heading

You’ve used the same connector for two different jobs in 24 hours. That’s the pattern: one connector unlocks dozens of briefs. Calendar is next.

Tomorrow on Switched On

Day 9: Walk into next week like you’ve had a chief of staff overnight.

Sunday edition. Connect Claude to your calendar. Eight minutes of prep on the three meetings of next week that matter most. Walk into Monday already two steps ahead.

Nishma x

P.S. How did today land? Three buttons, thirty seconds → Tap one

I read every response. The data shapes the rest of the challenge.

Catching up? Days 0 to 7 are on the Switched On Substack archive HERE.

​

Got a friend or a colleague who this might be useful for? Send her to nishma.co/30daychallenge

SwitchedOn #SwitchedOnAIChallenge

​

Welcome to Day 7 of the challenge.

On Day 5 you ran your bank statement through Claude Cowork and found subscriptions you’d forgotten. Cancelled some, kept some, flagged some as unsure. Money back.

You also hit a wall. The statement is brilliant at monthly subscriptions and useless at the annual ones. Apple, PayPal, Stripe, Amazon: they aggregate so many services that “APPLE.COM £79.99” tells you nothing about what you’re actually paying for.

The receipts that DO tell you live somewhere else. Your inbox.

Today we connect Claude to your inbox and let it sweep for what the statement couldn’t see.

A word before we start

This is your first connector. The shift is real.

Until now, Claude has worked with what you give it: typed text, attached files, your Cowork folder. Connectors flip that. You grant Claude access to a service (Gmail today, Calendar later, Drive later), and Claude can read it directly. No more copying, pasting or downloading.

That’s the unlock that takes Cowork from “useful assistant” to “actual colleague.” It’s also where Day 4’s permissions teaching earns its keep. Read-only first. Specific resources only. Disconnect when done. Same rules.

A note on Gmail specifically: if your primary inbox isn’t Gmail, the Outlook and other connectors work the same way. Skip the Gmail-specific clicks below and find the equivalent in your provider. Same brief, same payoff.

Step by step

Step 1. Open Cowork in your desktop app.
Click Customise or Settings.
Find Connectors (or Integrations depending on which version you’re on).

​

Step 2. Click Gmail. Sign in with your Google account.

Step 3. When Google asks what Claude can access, grant READ-ONLY access only.
​Not “Read, compose, send, and permanently delete.” Just ‘Read-only’.
If the screen offers a choice, untick anything that lets Claude send mail, delete mail or modify your account settings.
Read-only is enough for today’s task and for almost everything you’ll ever ask Claude to do with your inbox.

If you only see “all of Gmail” as the access option, that’s fine for today.
We come back to label-level scoping (only granting access to specific Gmail folders) when we go deeper on connectors.

Step 4. Open your Day 5 Cancel-Subs table from your Cowork folder. Keep it visible.

Step 5. In a new Cowork chat, paste this brief:

Search my Gmail for emails from the past 13 months containing any of these phrases: “subscription”, “renewal”, “trial”, “thanks for subscribing”, “your payment”, “auto-renew”, “annual renewal”, “your receipt”.
​

From the results, identify every recurring subscription service. For each one, list: company name, what the service is, how often it bills (monthly, annual, quarterly), and the amount if visible.
​

Cross-check against the attached table from my Day 5 bank statement analysis. Tell me what’s in my inbox that’s NOT on the table. That’s what I missed.
​

Format as a clean list grouped by frequency. No em dashes, no double-hyphens in your reply.

Step 6. Attach your Day 5 Cancel-Subs table. Send.

Step 7. Claude searches your inbox, cross-checks, and gives you a list of the subs the statement missed. Usually annuals. Often three to five new ones. Sometimes more.

Step 8. Go through the new list. For each one:

• Keep. Still using it.

• Cancel. Forgot about it.

• Unsure. Need to look it up.

Add them to your Day 5 table.
Or my preferred option is do this inside Cowork.
Just say: “Add a Status column. For each row, ask me what I want to mark it as, one at a time.”​
​

Step 9. For each “Cancel,” use the same Day 5 brief:

What’s the cleanest way to cancel [service name]? Give me the link or the path. Don’t make me search for it.

Cancel as you go, or batch into a “to cancel this week” list.

Step 10. Save the updated table back to your Cowork folder.

Step 11. Turn the table into a live dashboard.

In the same chat, paste:

Build me an interactive HTML dashboard from the subscriptions table we just made. Single file I can open in any browser. No internet needed, no dependencies.
​

At the top: three big totals that update live as I change things.

1. Total monthly cost of everything marked “Keep.”
2. Total annual cost of everything marked “Keep,” with monthly costs multiplied by 12.
3. Projected annual saving from everything marked “Cancel.”

​
Below: the list of subscriptions as rows or cards. Each one shows the service name, what it is, frequency (monthly / annual / quarterly), amount, and a three-way status toggle (Keep / Cancel / Unsure). Toggling updates the totals in real time.

​
Add a “+ Add subscription” button at the bottom. Clicking it adds a new row I can fill in. Include a small delete option per row.

​
Save my changes locally so they persist if I close the tab and reopen the file.

​
Style: clean and editorial. Dark background. Generous spacing. Big totals at the top, calm typography below. Easy to scan, not busy.

​
Save the file as 2026-XX-XX-subs-dashboard.html in my Cowork folder.

Claude builds it. Double-click the file to open it in your browser. Bookmark the tab. Next quarter, drop a fresh statement in, run the sweep, and ask Claude to update the dashboard with anything new.

This is the version of “tool” you’ll be building properly in Week 4. Today’s is your first one, in two minutes, without thinking about it.
​
​

A safety note

Granting Claude access to your inbox is a meaningful step. It’s also reversible.

What Claude can do with READ-ONLY Gmail access: read emails, search emails, summarise emails, extract information from them. That’s it.

What Claude CANNOT do with read-only: send emails on your behalf, delete emails, modify your inbox, change your account settings.

If at any point you want to revoke access:

• In Claude: Settings, Connectors, Gmail, Disconnect.

• In Google: myaccount.google.com, Security, Third-party apps with account access, find Claude, Remove access.

Both work. Either revokes immediately. Claude no longer reads anything new from your inbox the moment you disconnect. (Anything Claude already pulled into a previous chat is still in that chat history until you delete the chat.)

NEVER grant write access to your inbox unless you have a specific, well-understood reason. Read-only handles 95% of useful inbox tasks. Write access is rarely needed and creates risk.

What you’ll notice

The inbox catches what the statement couldn’t see. Often it’s the annuals: software you signed up for last year, a magazine subscription you forgot, an Apple One bundle, an Amazon Prime renewal, the gym membership you cancelled but didn’t actually cancel.

The cancellation total from today usually beats Day 5’s. That’s not because the statement was wrong. It’s because the statement only sees half the picture.

Cross-check is the move. Two sources for one truth.

Why this works

Aggregator merchants on a bank statement hide what they’re for. The renewal email never does. The email subject line is literally “Thanks for renewing your subscription to [actual brand].”

That’s why this works in a single sweep. The inbox holds the receipts. Claude reads them. You decide.

It’s also why connectors are the unlock people remember. Browser Claude is for thinking. Connected Claude is for doing.

Key takeaway

One connector. One brief. The annual subs your statement couldn’t see, surfaced and actioned. The Day 5 table, completed.

You now have a workflow you can run every quarter: bank statement + inbox sweep + aggregator check. Three sources, one decision pass, money back.

One week in. How’s it landing?

Quick favour before the weekend.

This is the end of Week 1. The hardest week of the challenge, if I’m honest. The setup days. The first proper task. The first connector. The bit where people decide whether AI is for them or not.

I want to hear how it’s going for you.

Whether you’ve done every day or you’re catching up. Whether you’ve had a wow moment or you’re still in the “this is fiddly” bit. Whether you’ve built something or you’re still finding your footing. All of it useful.

Three questions, if you’ve got a minute over the weekend:

1. What’s surprised you most so far? The thing you didn’t expect.

2. What have you actually built or done? Even tiny counts. Especially tiny.

3. What do you wish I’d told you on Day 0? What’s missing from the brief.

Hit reply on this email and tell me. Or DM me on Instagram. Or comment on the Substack post. They all come to me.

I’m reading every reply. I’ll fold what I learn into the rest of the challenge so the next two weeks land sharper for the women still finding their feet.

Where this is heading

You’ve crossed two thresholds in a week.
Browser to desktop. Statement to inbox. Both stay locked in.
Every wow day from here builds on both.

Tomorrow on Switched On

Day 8: Clear the inbox debt while the kettle boils.

Saturday edition. The personal mail you’ve been ignoring. The receipts. The booking confirms. The school forms. Same Gmail connector, different brief, twenty minutes between you and a backlog that’s finally handled.

Nishma x
​

P.S. How did today land? Three buttons, thirty seconds → Tap one
I read every response. The data shapes the rest of the challenge.

Catching up? Days 0 to 6 are on the Switched On Substack archive HERE.​

​

Got a friend or a colleague who this might be useful for? Send her to nishma.co/30daychallenge

SwitchedOn #SwitchedOnAIChallenge