Using Claude as an accountability partner at work: a setup for ADHD brains

Something has started coming up in my consulting room. More than once recently, people I work with, capable professionals with ADHD who are quietly drowning in the administrative half of their jobs, have mentioned that they had begun using an AI assistant to plan their working day. None of them had read about it anywhere; they had each arrived at it independently. And the sentence I keep hearing is a version of the same one: it is the first thing that has helped without becoming another thing to maintain.

That got my attention, because "it became another thing to maintain" is precisely how most productivity systems die for people with ADHD. So this piece is an attempt to take the idea seriously: why it might genuinely suit an ADHD brain, where the evidence actually is and is not, what to be careful about (particularly if you work somewhere regulated), how it compares with dedicated tools like Motion and Trello, and how to set it up.

Three things to be clear about first

• There is no evidence base for AI as an ADHD coach. The underlying mechanisms are well evidenced, and I will cite them below, but "use a chatbot to plan your day" has not been tested in trials. This is a reasoned suggestion, not an evidence-based intervention, and I would rather say so plainly.
• This is a productivity aid, not treatment. It is not therapy, not a substitute for assessment, and not a replacement for medication where that is appropriate. An AI cannot notice that you are depressed rather than disorganised.
• Mind what you type. Confidentiality is the single biggest practical risk, and I have given it its own section below. If you work in finance, law, medicine or any regulated field, read that part before you start.

In a hurry? The short version

• Check your employer's policy first. Many regulated firms allow only an approved enterprise tool. Never paste client-identifiable or price-sensitive information. Plan in generic terms: "finish the quarterly commentary", not its contents.
• Why it may suit an ADHD brain: it does the sorting step that usually kills a plan. You can hand it an unsorted, panicky paragraph and get back an ordered list, rather than having to arrive already organised, which is the executive function you do not reliably have.
• It also manufactures "if X, then Y" plans on demand, which is one of the better-evidenced ways to turn an intention into an action, and it gives immediate feedback, which an ADHD brain responds to far more than a distant deadline.
• Against Motion and Trello: Motion is genuinely good at auto-scheduling and rescheduling but costs around $19 a month and still needs structured input. Trello is passive, so a neglected board becomes a monument to what you are not doing. Use the assistant for thinking and planning, and let your calendar hold the commitments.
• The rhythm matters more than the wording: a plan in the morning, a nudge in the middle, a review at the end. Six copy-and-paste prompts are below, including a standing brief and an unsticking prompt.
• Start with one prompt, not six, and attach it to something you already do, such as your first coffee. Expect it to go flat eventually; that is habituation, not failure, so change the prompts.
• Honest limits: there is no evidence base for AI as an ADHD coach, only for the mechanisms underneath it. It is not therapy, it will not make you do the thing, and it will not notice that the real problem is burnout rather than disorganisation.

Why this might suit an ADHD brain

ADHD is, at its core, a difficulty with executive function: the machinery of planning, prioritising, starting, holding a goal in mind and monitoring progress. The most reliable compensation, and the one Russell Barkley has argued for consistently, is to externalise that machinery, to move it out of your head and into the environment at the point where the action is needed.1 Lists, alarms, whiteboards and calendars all work on this principle. A conversational assistant is an unusually good externalising surface for four reasons.

• It removes the sorting step, which is where ADHD plans usually die. Traditional tools require you to arrive with a structured task, correctly categorised, with a project and a due date. That structuring is the executive function you do not reliably have. With an assistant you can dump an unsorted, panicky paragraph and ask it to turn that into an ordered list. The tool does the part your brain finds hardest, rather than demanding it as the price of entry.
• It manufactures implementation intentions. Turning a vague goal into a specific "when X, I will do Y" plan has a medium-to-large effect on follow-through, with a pooled effect size of d = 0.65 across 94 studies.2 Asking an assistant to convert your list into time-anchored, concrete first actions is essentially an implementation-intention generator, and it will do it as many times as you need without sighing.
• It gives immediate feedback. ADHD brains respond to the immediate over the important-but-later, a pattern linked to differences in the dopamine reward pathway.3 A check-in that responds within seconds, and reflects your progress back at you, closes a loop that a static task list leaves open.
• It lowers the social cost of admitting you are stuck. A great many capable professionals will not tell a colleague they have been avoiding an email for three weeks. They will tell a machine. Whether or not you call it accountability, the honesty is what makes the plan usable, and shame is the usual obstacle to it.4

There is also a plausible relationship to body doubling, the practice of working alongside another presence, which many people with ADHD find helps them start. I would be cautious about leaning on that comparison, though: as I have written

elsewhere

, the controlled evidence for body doubling is thin and mixed even with a real human, and an AI is not a human presence. Treat any "it feels like someone is with me" effect as a bonus, not the mechanism.

The confidentiality problem, which comes first

This matters more than anything else in the article, and it is why I have put it before the practical setup rather than in a footnote.

• Do not paste confidential, client-identifiable or price-sensitive information. If you work in finance, that includes inside information and anything material and non-public. If you work in health or law, it includes anything identifiable about a patient or client. Plan in generic terms: "finish the quarterly commentary" rather than the contents of the commentary; "call the client about the restructuring" rather than who and what.
• Check your employer's policy before you start. Many regulated firms have an approved enterprise AI tool and prohibit consumer accounts for work purposes. Using a personal account for work planning may itself breach policy even if nothing confidential is typed. This is a five-minute question to ask, and much better asked first.
• Prefer the enterprise or work-provided version if one exists, which will normally come with the data handling your employer has already signed off.
• Keep personal health information out of it too. Notes about your ADHD, medication or mood are yours; think about where you would be comfortable with them living.

None of this makes the approach unusable. It just means the assistant knows your structure and not your secrets, which turns out to be all it needs for this purpose.

How it compares with Motion, Trello and the rest

Dedicated tools do real things that an assistant does not, and it is worth being fair about the trade.

• Motion

auto-schedules tasks onto your actual calendar based on deadlines and priorities, and reshuffles automatically when things slip. That automatic rescheduling is genuinely valuable for a brain that struggles to re-plan after a disrupted day. The trade-offs are cost, at roughly $19 a month or more, and that it still requires you to enter tasks in a structured way. Several people I have spoken to found the calendar filling up with auto-scheduled blocks they then ignored, which produces its own guilt.

• Trello

is visual, free at the basic tier, and excellent for seeing the shape of a project at a glance. Its weakness for ADHD is that it is entirely passive: the board never asks you anything, so a neglected board becomes a monument to what you are not doing.

• An assistant

such as Claude is conversational rather than structural. It is much better at the messy front end, taking an unsorted brain dump and producing a plan, at deciding what matters when everything feels urgent, and at unsticking you when you cannot start. It is worse at persistence, reminders and calendar integration, unless you deliberately give it a place to keep state. The free tier is enough to try everything in this article.

The honest conclusion is that these are complementary. The pattern I would suggest is: use the assistant for thinking, planning and reviewing, and let a calendar hold the commitments that come out of it. If you already have a task manager you actually use, keep it. If you have bounced off four of them, starting with conversation rather than structure is a reasonable thing to try.

The setup: prompts you can copy

Below is a working set. Adapt the wording, and note that the specifics matter less than the rhythm: a plan in the morning, a nudge in the middle, a review at the end.

1. The standing brief, which you set up once

Start a dedicated project or conversation for work planning, and give it a persistent instruction so you do not have to re-explain yourself daily. In Claude this can go in a Project's custom instructions; in other tools, look for "custom instructions" or "system prompt".

You are my work accountability partner. I have ADHD. My executive function is unreliable, particularly starting tasks, prioritising when everything feels urgent, and finishing administrative work that is boring but important.

How I want you to work with me:

Be warm, direct and practical. Never shame me for what I have not done, and never be falsely cheerful about it either. Treat undone tasks as information, not moral failure.

Always break tasks down until the first step takes two minutes or less, and phrase it as a physical action ("open the spreadsheet"), not an outcome ("do the expenses").

When I give you an unsorted list, sort it for me. Ask me at most two clarifying questions, then commit to a recommendation. Do not give me a menu of options to choose between, because choosing is the part I find hard.

Hold me to no more than three "must do" items a day. If I try to add a fourth, tell me what I am implicitly dropping.

I will not share confidential or client-identifiable information, so work with the structure I give you and do not ask for detail I have withheld.

2. The morning plan, five minutes

Here is everything in my head this morning, unsorted: [dump it all, messy is fine, do not tidy it first]

I have [X] hours of uninterrupted time today and these fixed commitments: [meetings].

Give me: the three things that genuinely must happen today, in the order I should do them; a two-minute first action for each; one thing on this list I should drop or delegate; and the single easiest win to start with, because I need momentum before I can face the hard one.

That last request is deliberate. Conventional advice says do the hardest thing first. For an ADHD brain the opposite is usually better: a small completed thing makes the next one easier to start.

3. The unsticking prompt, for when you have been staring at it

I have been avoiding [task] for [length of time] and I cannot make myself start. Do not give me a pep talk.

Ask me three short questions to work out what is actually in the way (is it unclear, is it boring, am I afraid of getting it wrong, is it too big, am I waiting on someone). Then give me one physical action that takes under two minutes, and nothing else.

4. The midday check-in, one minute

Midday check-in. Done so far: [list]. Not started: [list]. I have [X] hours left.

Re-plan the rest of the day realistically. Tell me plainly what is not going to happen today so I can stop carrying it, and what the one thing is that I should protect.

5. The end-of-day review, three minutes

End of day. Here is what happened: [what you did, including the unplanned things].

Reflect back what I actually achieved, including the things that were not on the list, because I tend not to count those. Note one thing that got in the way and might be worth changing. Then give me tomorrow's single first action, and stop there.

The "count the unplanned things" instruction matters more than it looks. People with ADHD routinely finish a day of genuinely useful reactive work and record it as a failure because it did not match the morning's list.

6. The weekly review, fifteen minutes on a Friday

Weekly review. This week I completed: [list]. Still outstanding: [list].

Ask me: what has been sitting on this list for more than two weeks, and what should now be either done, dropped, or delegated. Point out any pattern you can see in what I keep avoiding. Then help me pick the three things that matter most next week, and one thing I am going to formally abandon without guilt.

The permission to abandon something is not a throwaway. Undone tasks accumulate a psychological weight out of all proportion to their importance, and deliberately closing one is often worth more than completing two.

Making it stick, which is the actual hard part

• Attach it to something you already do. The morning plan happens when you sit down with your first coffee. An unanchored new habit will not survive a fortnight.
• Start with one prompt, not six. The morning plan alone is the highest-value one. Add the end-of-day review only once the morning one is automatic.
• Keep it in one place. A single ongoing project or conversation, so the context accumulates and you are not starting cold each day.
• Let the calendar hold the commitments. Anything with a time attached should end up in your actual calendar, because that is the thing that will interrupt you.
• Expect to fall off, and make restarting free. You will miss three days. The correct response is to open it on day four and plan that day, not to reconstruct the backlog or start a new system.
• Rotate when it goes flat. ADHD brains habituate. When the novelty wears off and the prompts feel stale, change them. That is maintenance, not failure.

What it will not do

It will not make you do the thing. Nothing does. It will not notice that the reason you have not written the report for a month is that you are burnt out, frightened of the feedback, or depressed, and it will cheerfully help you optimise your way around a problem that needs a different kind of attention altogether. It has no stake in your life, so the accountability is real only to the extent that you decide to treat it as real.

And it is worth saying that the tone you use with it tends to mirror the tone you use with yourself. If your inner voice treats every missed task as evidence of a defective character, you will build a system that quietly reinforces that, however clever the prompts. That is worth addressing directly, and I have written about it in

the ableist inner critic

. The tool works far better on top of a kinder standard.

This article is general information, not medical, legal or professional advice, and it cannot replace an assessment by a GP or specialist. Please check your employer's policy on AI tools and confidentiality before using any assistant for work, and follow it. I have no commercial relationship with any tool mentioned; pricing and features change frequently, so verify current terms yourself. If the underlying difficulty is affecting your work and wellbeing, that is a good reason to talk to someone, and the

free 10-minute consultation

is one place to start.

References and further reading

• Barkley RA. Executive Functions: What They Are, How They Work, and Why They Evolved. New York: Guilford Press; 2012. ISBN 978-1-4625-0535-7. On externalising executive function into the environment as the core compensation strategy.
• Gollwitzer PM, Sheeran P. Implementation intentions and goal achievement: a meta-analysis of effects and processes. Advances in Experimental Social Psychology. 2006;38:69–119.

doi:10.1016/S0065-2601(06)38002-1

• Volkow ND, Wang G-J, Newcorn JH, et al. Motivation deficit in ADHD is associated with dysfunction of the dopamine reward pathway. Molecular Psychiatry. 2011;16(11):1147–1154.

doi:10.1038/mp.2010.97

• Beaton DM, Sirois F, Milne E. Experiences of criticism in adults with ADHD: a qualitative study. PLOS ONE. 2022;17(2):e0263366.

doi:10.1371/journal.pone.0263366. On internalised criticism and shame in ADHD.