🧠 AI That Works With Your ADHD, Not Against It

I have ADHD, and I have a drawer’s worth of dead productivity systems to prove it. The bullet journal with nine beautiful pages and 200 blank ones. The app with the streaks (the streak died on day eleven and took my self-esteem with it). The colour-coded wall planner I stopped seeing after a week, the way you stop seeing your own wallpaper.

Here’s what took me embarrassingly long to figure out: those systems weren’t failing because I lacked discipline. They were failing because every one of them charged an entry fee my brain couldn’t pay - open the app, remember the system, categorise the task, maintain the streak. Neurotypical systems assume the hard part is doing the task. For brains like mine, the hard part is the three seconds before the task, and any system that adds steps to those three seconds is already dead.

Which is why AI is the first thing that’s actually stuck. Not because it’s clever - because a chat window has no entry fee. No categories, no setup screen, no streak to protect. You type the mess as it exists in your head, and something useful comes back. These are the five setups that have survived contact with my actual brain. I call them the Low-Friction Five, and they work precisely because they’re almost nothing.

1. The body-double chat

What it is: Body doubling is the ADHD classic: things get done when someone’s just… there. A friend on the sofa while you sort the paperwork. This is the AI version - a chat that sits with you through a task, checking in, keeping you tethered.

How to do it: Open a chat, paste in the body-double prompt below, and tell it the task. Then keep the chat open while you work and report back like you would to a friend: “done the first bit”, “got distracted, found myself reading about the Roman Empire”, “ugh, this form”.

Nick this one:

Be my body double while I do a task I've been avoiding. The task: [what it is]. I have ADHD, so here's your job: ask me what the very first physical action is and wait for me to say I've started. While I work, I'll message you when I finish bits or when I drift. When I drift, don't lecture me - just cheerfully point me back to where I was. Every so often ask "what's happening now?" if I've gone quiet. Keep replies to a couple of lines, warm but not gushing. Never tell me the task is easy. If I say I'm stuck, ask me what the next two-minute piece is. We're done when I say done, and then you can be pleased with me.

The non-obvious bit: The AI’s replies aren’t what’s doing the work - it’s that reporting to someone makes the task socially real. Also: on the phone app, voice mode makes this eerily like having a mate in the room while you clear the kitchen. Yes, it works even though you know it’s not a person. Brains are funny like that.

2. The task-initiation prompt

What it is: For the wall between you and starting. Not the whole task - just the first two minutes of it, handed to you so small there’s nothing left to negotiate with.

How to do it: Save this somewhere instant (notes app, or pinned in Claude), and fire it at anything you’ve been circling for days:

Nick this one:

I've been avoiding this task: [task]. I have ADHD and the problem is starting, not doing. Don't give me a plan, don't give me steps 1-10, don't explain why it matters - I know why it matters, that's why it hurts. Just tell me the first two minutes: the literal physical actions, like "stand up, get the shoebox of receipts, put it on the table, open it". Then stop. If I come back and say "done", give me the next two minutes only. Keep going in two-minute pieces until I tell you I've got momentum.

The non-obvious bit: The “then stop” matters more than anything else in there. A full plan re-triggers the overwhelm that caused the avoidance; two minutes is beneath your brain’s threat threshold. Say the task is “sort out the car insurance renewal” - three weeks of dread, and the first two minutes turn out to be “find last year’s email, open it, screenshot the price”. That’s it. That’s doable while the kettle boils, and starting is the whole war.

3. The brain-dump ritual

What it is: A regular emptying of the seventeen browser tabs in your head into one chat, where AI does the sorting your brain refuses to do. The thoughts don’t have to arrive in order, or in sentences, or make sense.

How to do it: Whenever the static gets loud (for me it’s most mornings, and Sunday evenings), open a chat and pour: “school shoes, that email to Dawn, car MOT is due?? cancel the trial before it charges, mum’s birthday, the thing I said I’d send Priya…” Then: “Sort this into: needs doing today, this week, some day, and things that are just worries, not tasks. Anything with a deadline first.” Done. The full ritual gets its own guide in Turn A Messy Brain Into A Plan.

The non-obvious bit: That last category - worries, not tasks - is the sneaky-powerful one. Half of what’s clogging an ADHD head isn’t actionable, and seeing it labelled as “not a task” releases it in a way no to-do list ever has.

4. Voice notes instead of typing

What it is: Talking your thoughts instead of typing them, using voice mode in the phone app. For a lot of ADHD brains (mine included), speech comes out before the self-editing kicks in - typing gives you time to delete the thought, judge it, or wander off mid-sentence.

How to do it: Phone app, tap the voice icon, and ramble while you walk, drive the school run, or lie on the floor (no judgement, floor time is valid). Ramble the brain dump, ramble the email you can’t start, ramble the “help me decide” spiral. Then ask for it back as a list, a draft, or a decision laid out with pros and cons.

The non-obvious bit: Movement plus talking is a genuine ADHD combination - some of my best untangling happens mid-walk with headphones in, looking like someone having a very animated phone call. The transcript doesn’t care that you said “um” forty times or changed subject twice. It catches everything and holds it.

5. The one-chat-not-twelve rule

What it is: A rule against the ADHD special: twelve open chats, each abandoned mid-thought, none findable later. You keep ongoing life-admin in ONE running chat (mine is literally called “brain”), and open fresh chats only for proper standalone jobs.

How to do it: Start a chat today, name it, and return to it for the daily dumps and check-ins. With memory switched on (Settings → Capabilities - and turn on “search and reference chats” while you’re there), Claude also remembers you across conversations, so even your abandoned tabs aren’t truly lost. Menus move around occasionally - poke about, it’ll be nearby.

The non-obvious bit: This is object permanence insurance. ADHD brains famously lose what they can’t see; one named chat means there is exactly one place your life-thread lives, and it’s always where you left it. Choosing where to put a thought is friction, and friction kills the habit - one chat means zero choosing.

What scaffolding can’t do

Said with love and lived experience: AI is scaffolding, not a cure. It won’t manufacture dopamine, it won’t make executive function appear on demand, and it’s not a substitute for the real supports - diagnosis, medication if that’s your route, therapy, humans who get it. And some days, nothing on this list will stick. Not the body double, not the two-minute trick, nothing. That’s allowed. A bad brain day isn’t you failing the system; the system waits, entry fee still zero, for whenever you wander back. That’s the entire reason this one survives where the bullet journal didn’t.

If you do nothing else: save the task-initiation prompt somewhere you can reach in five seconds. It earns its keep the very first time the wall goes up.

From one interesting brain to another

You don’t need more discipline, you need lower doors - and these five are the lowest I’ve found. For the genuinely rough days, the days when even two minutes is too many… there’s a gentler protocol built exactly for those: The AI Setup That Works When Your Brain Won’t. ✨