AI for Occupational Therapists

Updated January 2026 | 9 min read

A stroke patient needs adaptive equipment for dressing. You evaluated them three weeks ago. Their affected arm had 2/5 strength, minimal active range of motion, moderate spasticity. You recommended a button hook, elastic shoelaces, and a dressing stick. Today they're back. Their strength improved to 3/5, but they're still struggling with buttons. You need to decide if they need a different tool or more practice with the current one.

That decision requires comparing today's function to baseline. Your EMR has the initial evaluation scores. It doesn't capture the specific dressing challenges—the patient lives alone, has arthritis in the unaffected hand, wears clothes with small buttons for work. Those details change what equipment makes sense.

Standard AI tools don't hold this context. ChatGPT can list adaptive equipment options. It can't remember that this patient already tried a button hook and found it difficult to manipulate with their non-dominant arthritic hand. It doesn't know their goal is returning to their office job where elastic-waist pants aren't appropriate.

Functional Baselines That Drive Treatment

Occupational therapy focuses on function. Can the patient dress themselves? Prepare a meal? Manage medications? Return to work? These activities break down into component skills—fine motor control, cognitive sequencing, endurance, problem-solving.

You establish baselines at evaluation. The patient can button three buttons in two minutes with moderate difficulty. They can prepare a cold meal but not a hot one. They can remember to take morning medications but forget evening doses. These baselines are specific and measurable.

Progress is measured against these baselines. Four weeks later, the patient buttons five buttons in 90 seconds with minimal difficulty. They can now prepare simple hot meals using the microwave. Evening medication adherence improved after you set up a reminder system.

Your EMR stores these data points. AI with memory connects them into a progress narrative. You see not just improvement in scores, but improvement in real function. You track which interventions drove which changes.

Activity Tolerance and Energy Conservation

Many OT patients have limited endurance. Cardiac patients. COPD patients. Chronic fatigue. Post-viral syndrome. They can perform activities, but they run out of energy quickly.

You teach energy conservation techniques. Sit while cooking. Use a shower chair. Pace activities with rest breaks. Modify the home environment to reduce steps and reaching.

Progress here is subtle. The patient can now cook dinner without needing to lie down afterward. They can shower and dress in the morning and still have energy for other activities. They're doing more overall while experiencing less fatigue.

This progress doesn't show up in traditional outcome measures. It shows up in patient reports and functional observations. AI that remembers these details tracks the changes that matter most to quality of life, not just the changes that fit on assessment forms.

Adaptive Equipment Trials and Outcomes

You have a closet full of adaptive equipment. Reachers, sock aids, weighted utensils, built-up handles, jar openers, one-handed cutting boards, elastic shoelaces, long-handled sponges. You know what each tool does. What you learn through experience is which patients will actually use which tools.

The reacher seems like an obvious choice for someone who can't bend down. But if the patient has poor grip strength, they can't operate the reacher effectively. The weighted utensils help with tremor, but they increase fatigue for patients with weakness. The sock aid works well if the patient has good cognitive function to learn the technique, but it's frustrating for someone with memory deficits.

You trial equipment and track what works. This patient did well with the button hook after practicing for two sessions. That patient couldn't master the sock aid but succeeded with compression stockings. Another patient rejected all dressing aids but accepted clothing modifications instead.

AI with memory tracks these trials. When a new patient presents with similar impairments, you know which equipment is worth trying based on past success patterns. You don't waste time on tools that rarely work for patients with that specific combination of limitations.

Home Safety Assessments and Modifications

You do home visits or detailed interviews about home layout. Where are the stairs? Is the bedroom on the first floor? How many steps to enter the house? Is the bathroom accessible from the bedroom? Are there grab bars? Rugs that could slip? Adequate lighting?

Safety recommendations depend on specific home features and patient deficits. A patient with balance problems needs grab bars in the bathroom, but grab bar placement depends on toilet and shower configuration. A patient with low vision needs better lighting, but the type of lighting depends on existing fixtures and room layout.

You track what modifications you recommended and whether the patient implemented them. You note barriers to implementation—cost, landlord restrictions, family resistance. When the patient returns, you need to know what changes were made so you can assess whether safety has improved.

Your EMR might note "recommended grab bars." AI with memory notes "recommended grab bars next to toilet on left side and inside shower stall on back wall. Patient's son will install within two weeks. Follow up at next visit to confirm installation and assess patient's ability to use them safely."

Cognitive and Perceptual Impairments

Many OT patients have cognitive deficits. Stroke. Traumatic brain injury. Dementia. These impairments affect their ability to learn compensatory strategies.

You assess attention, memory, problem-solving, sequencing, safety awareness. A patient with memory deficits needs external aids—checklists, labeled cabinets, medication organizers. A patient with sequencing problems needs tasks broken into smaller steps. A patient with poor safety awareness needs environmental modifications and caregiver training.

Progress in cognitive rehab is gradual. The patient who couldn't remember a three-step task now handles four steps with a written list. The patient who left the stove on twice last month has used a timer successfully for three weeks. The patient who couldn't organize their medication now uses a pill organizer independently.

These improvements don't fit standard assessment tools well. They're functional changes that emerge from session-to-session observations. AI with memory tracks the specific strategies you've implemented and how well each patient is using them.

Training Caregivers

Many OT patients need caregiver assistance. You're training family members or paid caregivers on safe transfers, cueing strategies, how to set up tasks for success, when to step in versus when to allow independence.

Different caregivers learn at different rates. Some pick up techniques immediately. Others need repeated demonstration. Some are overprotective and do too much for the patient. Others underestimate safety risks.

You document caregiver training in your notes, but the nuances matter. This caregiver masters physical techniques quickly but struggles with knowing when to cue versus when to wait. That caregiver is great at following your instructions but can't problem-solve when something unexpected happens. This caregiver is reliable but only available three days a week, so the patient also needs strategies for the other days.

AI that remembers caregiver strengths and limitations helps you design training that works. You focus your teaching on each caregiver's learning needs and design the care plan around who's available when.

Work Rehabilitation and Return-to-Work Planning

A patient's goal is to return to their job. You need to know what the job requires. An office worker needs computer skills, sustained sitting, cognitive endurance. A construction worker needs strength, balance, coordination, ability to work at heights. A teacher needs standing tolerance, fine motor skills for writing, vocal endurance.

You assess job-specific skills. Can the patient type for the duration their job requires? Can they lift the weights they'll encounter at work? Can they stand for their expected shift length? These assessments guide treatment and return-to-work timing.

You also work with employers on accommodations. Modified schedules. Ergonomic equipment. Job restructuring. These negotiations require knowing both the patient's functional capabilities and the employer's flexibility.

AI with memory tracks job requirements, functional progress toward those requirements, and accommodation options you've explored. When it's time to write a return-to-work letter or meet with the employer, you have all the relevant information ready.

Pediatric Development and School-Based Goals

Pediatric OT addresses developmental delays and school function. Handwriting. Scissors use. Self-care skills like dressing and toileting. Sensory processing. Social participation.

Progress happens over months and years. A five-year-old who couldn't hold a pencil correctly now writes their name legibly. A seven-year-old who melted down during school assemblies now participates with noise-canceling headphones. A ten-year-old who couldn't tie shoes learned an alternative technique that works for them.

You're coordinating with parents, teachers, and sometimes other therapists. Each adult in the child's life sees different aspects of function. The teacher sees classroom performance. The parent sees home and community function. You see the underlying skills that enable both.

Tracking progress across settings and coordinating between adults requires holding a lot of context. AI with memory keeps the full picture—what's working at school, what's challenging at home, what strategies each adult is implementing, what the child's next developmental targets are.

Sensory Integration and Regulation

Some patients have sensory processing issues. They're over-responsive to noise, light, touch, or movement. They're under-responsive and seek intense input. They have difficulty regulating their arousal level—they're either overwhelmed or shut down.

You identify sensory patterns through observation and interview. This child covers their ears in loud environments. That adult can't tolerate certain clothing textures. This patient seeks deep pressure and benefits from weighted items. That patient needs movement breaks to maintain attention.

Treatment involves environmental modifications and sensory strategies. Noise-canceling headphones. Fidget tools. Weighted blankets. Movement breaks. Textured materials for tactile seekers. Dimmed lighting for visual sensitivity.

What works is highly individual. You trial different strategies and track results. AI that remembers which sensory strategies helped which symptoms lets you refine the sensory diet over time. You build on what works and eliminate what doesn't.

Insurance Documentation for DME and Services

Durable medical equipment requires insurance authorization. Wheelchairs. Hospital beds. Patient lifts. Shower chairs. These items are expensive, and insurance companies want documentation of medical necessity.

You provide that documentation. You describe the patient's functional limitations. You explain why this specific equipment is necessary. You document that less expensive options are inadequate. You include measurements, assessments, and clinical justification.

Different insurance companies have different requirements. Medicare wants specific functional limitation codes. Medicaid has different approval processes by state. Private insurance often requires prior authorization before evaluation.

Keeping track of these requirements per payer keeps your authorization requests from getting delayed or denied. AI with memory holds insurance-specific documentation requirements. You know what information each payer needs before you write the request.

The Setup for Occupational Therapy Practice

One markdown file. CLAUDE.md sits in a secure location on your device. Inside that file: patient functional baselines, equipment trials and outcomes, home modification recommendations, cognitive strategies per patient, caregiver training notes, work/school requirements, sensory patterns, insurance documentation requirements.

Every time you use Claude Code, it reads that file first. The AI knows your patients' progress, the equipment they're using, the strategies you've implemented. You don't re-explain why this patient needs a specific adaptation. You don't look up insurance requirements for the fifth time this month.

When you learn something—an equipment trial that worked well, a successful strategy for a challenging case, an insurance quirk—you add it to the file. Your clinical knowledge accumulates.

No EMR integration. No complex software. One file gives AI persistent memory across every session.