What can AI do for OT practice, and what stays with you?

That's where AI is worth a look for OTs. Not in the assessment. In the writing that comes after it.

What there isn't yet is good UK evidence. Most of what's published about AI for OTs is American, written by companies selling documentation tools to US practices. So this page isn't a review of proven OT tools. It's a map of where OT work gets stuck, where AI might help, and the same information governance rules that apply everywhere else in Clinically Intelligent.

Why OT documentation takes so long

Environmental assessments. The visit is the quick part. The report takes far longer. I measure the space. I describe what the person wants to be able to do. I work out what the home can take — whether the walls are suitable for rails, whether the bathroom can be adapted, whether the stairs suit a stairlift. I set out each recommendation and the reasoning behind it, alongside the person's functional level. All of it has to make sense to someone who wasn't in the room — the family, a contractor, whoever approves the funding.

Social histories. On the ward, this is the one I like least. Who the person lives with. What help they already have. What their home is like. What they could do before they came in. It's a long list of questions, a lot to hold in your head, and all of it written up afterwards.

Comprehensive geriatric assessments. A CGA looks at the whole person: physical health, function, mood and cognition, social circumstances, environment. It's long by design. You want it detailed, with every piece of information there, because the decisions that follow depend on it.

Even a standard initial assessment takes longer to write than it should.

That's the thing about OT documentation. It's long because it has to be. A missing detail can mean the wrong equipment, a plan that doesn't get approved, or a discharge that doesn't hold. The aim isn't to write less. It's to spend less time turning what you found into words.

Where AI can actually help

The documentation. Everything above, plus intervention plans and discharge paperwork. A lot of it follows a set structure, which is exactly the kind of writing AI is good at drafting. Anything that could identify a client stays out of public tools. Draft from anonymised or made up details, then personalise it inside your Trust's systems.

Organising your caseload. Waiting lists, prioritising, chasing referrals, remembering which review is due when. The admin around the clinical work, the bit that eats the end of the day.

Explaining things to patients. Turning a rehab plan into something a person can actually follow at home. You find the evidence, test it against the questions people actually ask, then explain it plainly. AI can help with that last bit, the explaining. The clinical judgement stays with you. I walked through a full version of this in Issue 04, building a carer guide for someone recovering from a brain injury.

OTs already turn clinical reasoning into plain language every day. It's a big part of the job. AI can write it faster, but knowing what this particular person needs to hear, and how to say it, is still you. The tool does the sentences. You do the sense.

The work around the work. CPD, reflective accounts, service development. The stuff that never gets enough time. In Issue 09 I used Gemini to help me reflect on my own work for CPD, and it picked up a gap in my thinking I hadn't spotted.

One thing to try this week

Pick the piece of documentation you dread most, the one you leave until Friday. Next time, write your usual headings and bullet points first, because the clinical thinking is yours. Then ask an AI tool to turn the bullets into proper sentences. Read every line and correct it. Notice what it got wrong. That's the interesting bit, because it shows you where your own notes aren't as clear as you thought.

Ten minutes. One document. You're not signing up to a new system. You're just testing whether it earns a place in your week. If the draft is worse than what you'd normally start with, you've learned something for free. If it's better, you've got your Friday afternoon back.

If you want help asking for a better draft, the AI Prompting Guide walks you through it. It's free.

What doesn't change

Same rules as always. Nothing that could identify a client goes into a public AI tool: no names, no dates of birth, no mix of details that could point to someone. Anything AI writes for clinical use gets checked by you before it goes anywhere. And if your Trust hasn't approved a tool for something, don't use it for that until it has.

If you want the full list, the Clinician's AI Safety Check is nine questions to run before you use AI at work. It's free.

Before you rely on this

This is a starting map, not an evidence review, because there's no independent UK evaluation I can point you to of the documentation tools being sold to OTs. It's too early to recommend specific tools. Nearly all of them are American, and most of what you'll read about them comes from the people selling them. I'll update this page as UK pilots and evaluations come out.