I scored 147 on the RAADS-R and 38 on the AQ-50 - both autism screening tests, not ADHD. But I have ADHD too. The combo means my brain is running two different operating systems at once, neither of which came with a manual. I've known for years that my brain works differently, but seeing the numbers made something click.
How ADHD and AI work together - that's become the actual question I've spent the last year building answers to. Not "how do I fix my ADHD" and not "how do I force myself to be more productive." More like: where are the exact gaps, and can software patch them?
Short answer: yes. Not all of them. But enough to matter.
What ADHD actually costs (it's not what productivity gurus think)
The mainstream narrative is that ADHD means you're distracted and you need to focus better. That's a useless frame. The real problem is executive function failure - the part of your brain that handles task initiation, working memory, time perception, and emotional regulation around boring tasks.
Concretely, that looks like this for me:
- I forget things. Not occasionally. Constantly. Including things I cared about 20 minutes ago.
- I start projects in hyperfocus bursts and sometimes walk away mid-sentence when the dopamine dries up.
- I have a productive window from roughly 11am to 2pm. After 2pm I crash. Creative work comes back around 7pm. This isn't laziness - it's a real energy curve that I've tracked across years.
- Decision fatigue hits me hard. By 3pm, choosing between two options of roughly equal weight can genuinely stall me for 30 minutes.
- Calls destroy my flow state. I've worked async-only since I started freelancing at 14 because traditional structure never fit.
None of this is fixed by a better to-do list app. I've tried every productivity system. GTD, time blocking, Notion templates, physical planners - all failed because they assume linear execution, consistent motivation, and that you'll remember to check the system. ADHD brains don't do any of those things reliably.
What AI does is different. It meets you where you are instead of where you should be.
How ADHD and AI work together in practice
I run a multi-agent system called OpenClaw. One of the agents is borb - that's my day-to-day assistant, and it's the piece that's changed how I work more than anything else in the last two years.
The key insight: AI doesn't need you to have executive function. It has its own.
Borb maintains memory files. When I forget what I was doing - and I will forget - it has a record. When I say "remind me to follow up on this," it doesn't forget. When I ask the same question for the third time, it answers without judgment. That last part is underrated. Executive dysfunction comes with a lot of shame. An agent that just answers, every time, without the sigh or the "didn't we talk about this?" - that's genuinely useful.
The workflows that actually stuck:
Morning state load - borb checks my reminders, pulls calendar events, and gives me a plain-English summary of what's in front of me. I don't have to go hunting across four apps to reconstruct what today looks like. It's already there. This sounds small. It is not small. Starting the day without that reconstruction overhead means I hit my 11am-2pm window with actual energy instead of spending it getting oriented.
Building - I describe what I want in rough terms, the agent scaffolds the structure, and then I hyperfocus on the interesting parts. I've talked about this in building 7 apps at once - the ADHD pattern of "start fast, lose steam" becomes less destructive when the boring scaffolding is already done. The fun parts stay fun. The scaffolding doesn't require motivation.
Writing - AI drafts, I edit. Editing is dramatically easier than a blank page for an ADHD brain. The blank page has infinite options and zero constraints. A draft has something to react to. Even a mediocre draft gives me a direction to push against, and pushing against things is something my brain can do.
Research - This one might be the biggest one. I used to fall down 3-hour rabbit holes when I needed to look something up. Not because I was being inefficient - because that's what an ADHD brain does when it hits something interesting. Now I say "look into X and summarize it" and get back a usable synthesis in 10 minutes. The rabbit hole still exists. I'm just not in it.
Why ADHD and AI work together better than any productivity system I've tried
There's a reason GTD and every app built on top of it eventually fails for ADHD brains. The research on this is pretty clear - executive function deficits aren't a motivation problem, they're a neurological regulation problem. Systems that require you to consistently initiate, remember to check in, and maintain motivation to follow through are asking the exact thing the ADHD brain struggles with most.
AI sidesteps that loop in a way static systems can't. A to-do app is passive. It waits for you. An AI agent can be active - it surfaces things, drafts things, reduces the next action to something trivially small. That difference is the whole game.
The clinical picture backs this up too. Research from ADDitude Magazine's medical review board consistently shows that external scaffolding - someone or something that holds structure on your behalf - is one of the most effective accommodations for executive dysfunction. That's what ADHD coaches do. That's what a good EA does. AI can do a version of it at a fraction of the cost, 24 hours a day, without the social friction of asking a human for help again.
The difference between those external scaffolding approaches and AI: AI doesn't get tired of you.
The 90-day planning system I actually use
One of the more structured things I built is a multi-model planning system for quarterly work. I run it through Claude, Perplexity, and DeepSeek because they think differently.
Here's the rough structure:
# Run planning session across models
# Claude: strategic framing and goal coherence
# Perplexity: research-backed context on what's realistic
# DeepSeek: adversarial review - what did I miss, what's wishful thinking
Claude does the strategic layer - takes my rough "here's what I want to do this quarter" and shapes it into something coherent with actual priorities. Perplexity pulls external context - market stuff, comparable benchmarks, anything grounding. DeepSeek punches holes in it. Tells me where I'm being optimistic, where the plan assumes consistent motivation that I won't have.
The output is a quarterly plan that's actually calibrated to how I work, not how I should work. It accounts for the energy curve. It front-loads high-executive-function tasks to the 11am-2pm window. It builds in slack because I will start a side project mid-quarter.
This cost me maybe $3 in API calls. ADHD coaches charging $500/month are selling something similar - someone to help you structure the chaos. The AI version doesn't replace the human relationship, but it does the structural work well enough that I stopped needing the coach.
What still doesn't work
I want to be honest here because this is where most "ADHD and productivity" content fails - everything sounds like it works great.
Most AI productivity tools are built by neurotypical people for neurotypical people. They assume:
- You'll check the app on a schedule
- That once you've planned something, you'll follow the plan
- That motivation is a dial you can turn up
- That reminders are sufficient to trigger action
None of that is true for ADHD brains. A reminder I ignore three times isn't a reminder problem - it's a task initiation problem. The reminder sees the symptom and treats it by making more symptoms.
The AI tools that work for me are the ones that do the task or reduce it to the smallest possible action. The ones that fail are the ones that just tell me about the task more times. "You have 5 overdue items" is not useful. "I drafted an email for item 3, want me to send it?" is useful.
The other failure mode: AI planners that generate beautiful structured plans I'll never look at again. I've built these myself. They're impressive to screenshot, useless in practice. A plan I'll actually use is ugly, has a lot of slack built in, and survives contact with a day where I wake up unable to do hard things. Most AI-generated plans look like they were made for someone who executes flawlessly and never has a 2pm crash.
The design principle underneath all of this
The best automations I've built are invisible. They give me back the cognitive capacity I was spending on logistics, and I don't notice them working - I just notice I'm less exhausted.
That's the principle I use now when deciding whether to build a new automation: does this return brain capacity, or does it just offload a task I didn't care about?
Offloading boring tasks is fine. But the real value is when an automation handles something that was eating working memory - that background process running in my head that's tracking ten things at once. When borb maintains my memory files, I'm not the one holding context across sessions. That background process is freed up. That's where it gets actually useful.
The principle maps to AI broadly: don't use it to do more things. Use it to think fewer thoughts you didn't want to be thinking.
How I'd actually start if I were building this from scratch
If you're ADHD and you've read this far, here's the concrete version:
Pick one gap - not "I want to be more productive." Pick the specific thing that burns the most energy. For me it was morning orientation. For you it might be task switching, or research rabbit holes, or writing block.
Build the smallest intervention - don't automate your whole life. Build one thing that addresses one gap. If it sticks, expand it. If it doesn't, the cost was low.
Make it push, not pull - tools you have to check don't work for ADHD. Tools that surface things to you do. The difference between "you should open the app" and "here's your summary in Slack at 9am" is the difference between a system I use and one I abandon.
Use the right model for the task - I wrote more about my AI engineering stack in the tags section if you want to see the breakdown. Fast cheap models for logistics, more capable models for judgment calls. Don't run GPT-4-class inference for reminders.
Don't trust AI planning apps - build your own, even if it's rough. An n8n workflow or a cron job calling an API isn't glamorous but it works because you designed it around your actual workflow, not a generic one.
Start with memory before anything else - the single highest-leverage thing AI can do for an ADHD brain is hold context you'd otherwise lose. A markdown file your agent reads and updates is worth more than any reminder system.
The real unlock was accepting that I'm not going to become a different person with a different brain. The goal is building systems that work for this brain. AI is the first category of tool where that's actually been possible at the software layer.
If you're somewhere on the spectrum of neurodivergence and you've been grinding through systems that don't fit - this is worth trying. Not because AI fixes it. Because for the first time, the tools can be weird enough to match how you actually work.
The my coding stack I run is open in spirit if not fully in docs. If you want to talk about how to build something similar, I'm usually in the comments or findable.
Top comments (0)