masterclass · 4 hours

Practical AI for private practice

Kinatex education committee masterclass · Michael Rowe, University of Lincoln

§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Today

  1. Communication and education : patient handouts, exercise programmes, checking what AI gives you
  2. Professional development : research, guidelines, your own learning
  3. Administrative efficiency : Kinatex's rules, templates, everyday admin, an AI policy for patients
  4. Business operations : a persona, content in your own voice, your website, a new service

A question for the day: which parts of your work could AI take on, and which have to stay yours?

Bring a phone or laptop with a free AI tool open; a laptop makes uploads easier.

§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

AI now works alone for hours

  • About 7 minutes in May 2024; about 17 hours by April 2026 (METR, 2026)
  • Doubling roughly every 4 months since 2023 (METR, 2026)
  • If you last looked at the cutting edge a year ago, today's free tier is well past what you saw
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

AI now works at a clinical level

  • In simulated consultations, a diagnostic AI took histories and reached diagnoses at least as well as primary care doctors (Tu et al., 2025)
  • Doctors working with GPT-4 reached the same standard as GPT-4 working alone (Goh et al., 2025)
  • ChatGPT and DeepSeek answered 60 rehabilitation questions to a standard rated above that of final-year physiotherapy students (Michou et al., 2026)
  • AI replies were rated more empathetic than clinicians' in 13 of 15 studies (Howcroft et al., 2025)
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

AI is already improving care for real patients

  • Emergency department second opinions : the right diagnosis at triage in 66% of cases, where the two attending physicians on the same cases reached 54% and 48% (Brodeur et al., 2026a)
  • Primary care : 100 patients took a pre-visit history with an AI; the final diagnosis was on its list in 90% of cases (Brodeur et al., 2026b)
  • Nursing : an AI reading patterns in nurses' documentation; 35.6% lower risk of in-hospital death across 74 units (Rossetti et al., 2025)
  • Allied health private practice : 119 clinicians used an AI scribe for three months; less time on notes and letters, less charting after hours, and a better therapeutic relationship by clinicians' and patients' accounts (Evans et al., 2025)
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

On its own, the AI named the right condition 95% of the time. People using the same AI managed under 35%.

Bean et al., 2026

§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Why people with the AI did worse

  • They left out details that mattered : the AI can only reason from what it's told
  • They asked closed questions, e.g. "Could this be stress?" : a yes-or-no question limits what the AI can say back
  • They didn't pick the right answer when it was offered : choosing between two or three conditions takes clinical knowledge
  • The AI didn't always help them : it sometimes misread what it was told, or had the right answer and didn't make it clear

How you ask decides much of what you get back; that's what today practises.

§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

What this means for your practice

  • AI can already draft much of the writing, searching and planning around your clinical work, and on some clinical tasks it matches clinicians
  • Its output depends on the context it's given, and you have the context about this patient, this clinic, and this community
  • The judgement about what's right for the person in front of you comes from you
  • Today covers some of the work around that judgement: communicating, learning, administration, building a service
§02 · how AI works

How AI works

  • What it can see : the model works only from its training and what's in the conversation, which is why the context you give it matters
  • What it can't do : it doesn't learn from you, and it knows nothing about your clinic or your patients unless you tell it
  • Why it matters today : almost every habit in the four sessions follows from these two points
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

AI runs on tokens

  • Tokens : fragments of words; the model predicts the next one, then the next, over and over
  • Next-token prediction : enough to produce reasoning, planning and multi-step work
  • Training : done once, months ago, on an enormous body of text, producing a fixed model with an end date
  • Inference : what happens when you type; the model doesn't learn from you, and nothing you say changes it
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

AI sees only the conversation

  • Context window : everything the model can see at once, i.e. your prompt, its replies, anything you paste or upload (also, custom instructions)
  • Size : thousands of pages, e.g. your patient information pack, a clinical guideline and your policies all at once, so you can ask it to compare them
  • A new conversation : starts blank
  • "Memory" : a feature built on top; text is saved and handed back to the model
  • Long threads : the window fills up, and earlier material fades
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

AI can work with more than text

  • Images : a photo of a blank intake form, the whiteboard after a team meeting, a letter from an insurer
  • Audio : a voice note recorded after a long day, turned into a to-do list; or a spoken conversation in voice mode
  • Video : a clip of you demonstrating a home exercise, checked against the written instructions you give patients
  • It reads a screenshot of a web page as a page: layout, hierarchy, what catches the eye first
  • It's most useful when it interprets something real you've given it
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Cost keeps coming down

  • What needed a research lab's budget 18 months ago is on the free tier today
  • New capability arrives at the top end and reaches the free tier within months
  • Most of what you'll do today is possible on a free account; what limits you now is knowing what to ask for
  • A paid account is worth it when a limit stops you doing work you want to do, and not before

Epoch AI, 2025

§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Which tool: my experience

  • ChatGPT : the most familiar; a good all-rounder, with voice and image creation on the free tier
  • Claude : my choice for long documents and careful writing; tighter free-tier limits but still substantial
  • Gemini : free with any Google account; includes Deep Research on the free tier, capped
  • Copilot : may come with your clinic's Microsoft 365 account, possibly with data protections the others lack

The leading models are now close in raw capability; what separates them is the features, limits and data terms around them.

§03 · patient education

Session 1 ·
Communication and education

  • Patient education : the materials you give patients, i.e. handouts, home programmes, explanations of their condition
  • The method : start with a weak prompt and improve it step by step, then apply the same habits to an exercise programme
  • The check : compare what AI gives you against the standard you're accountable to
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Start simple

Write me a patient handout on low back pain.

  • Type this exactly as it is; typos don't matter
  • Read what comes back: would you give this to a patient on Monday?
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Give it the missing context

I'm a physiotherapist in a private clinic in Montreal. Most of the patients I see with low back pain are office workers aged 35–55 who've had recurring pain for over a year and have already tried rest and painkillers. Write a one-page handout that explains why movement helps, what to expect in the first few weeks, and when to come back to see me. Use plain language, warm but direct.

  • Same request as the first prompt, with the context around it
  • Swap in your own profession and patients, then compare the two outputs side by side (10 minutes)
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

What can be improved?

That's too generic; it could be for any clinic. What do you need to know about me and my patients to make it specific?

  • Treat the first response as a starting point
  • Telling AI what's wrong with a draft is a normal move; it's how you get to something usable
  • Notice what it asks for: who your patients are, what they've tried, what you want them to do next, how you want to sound
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Let AI interview you

Before we continue, ask me the 5 questions that would most change your answer.

  • You can use AI to build a good prompt; you don't need to know the exact shape in advance
  • Works on any task: a handout, a letter, a programme, a policy
  • If its questions seem off the point, your request was probably unclear; answer them anyway, or restate what you want
  • For bigger tasks, ask for one question at a time: you answer each properly, and each question builds on your last answer
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

The structure of a prompt

Role · Context · Task · Constraints · Output

You're an experienced clinician in musculoskeletal rehabilitation. (role)
I'm a kinesiologist in a Montreal clinic. My patient is 64, has knee osteoarthritis, and wants to keep hiking but is nervous about pain. (context)
Write a six-week home exercise programme. (task)
Twenty minutes a day, a chair and a resistance band only; include how to tell acceptable pain from a warning sign. (constraints)
Give it as a table: week, exercises, sets and reps, when to progress. (output)

§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Or ask AI to write the prompt

I want to [describe the outcome you want] but I'm not sure how to ask for it. Write the prompt for me, then tell me what you'd need to know to make it better.

  • Not every prompt needs all five; context and constraints are the two most often missing
  • Compare what it gives you with the five-part structure
  • Knowing the structure is how you judge whether its prompt is any good
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Your own exercise programme

The aim: practise the five-part structure on the kind of patient you see every week (10 minutes).

  • Invent a patient typical of your caseload: a condition you see often, their age, their goal, what worries them
  • Write a prompt for the programme you'd give that patient
  • Ask for dosage, progressions, and the signs that mean the patient should stop and get in touch
  • Then ask:

What would you need to know about this patient to make this safer?

§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Check it against the guideline

The aim: AI drafts from everything it has read, good and bad. A guideline is the standard you're accountable to, so use it as the reference and make the AI show where its draft departs from it.

Compare this programme with [named guideline] for [condition]. List every place it departs from the guideline, and quote the guideline's wording for each.

  • What to look for: outdated advice, invented references, confident claims with nothing behind them, dosage that's too aggressive or too timid
  • You decide whether to change the output or set it aside
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Do's and don'ts

  • Do give AI context about you and your patients before asking for output
  • Do invite clarifying questions
  • Don't put in anything from a real patient's record; invent a typical patient instead (more in Session 3)
  • Don't accept the first output as final
  • Don't assume it knows local guidelines, regulation or insurer requirements unless you tell it
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Discussion

  • Where did the output land better or worse than you expected?
  • Which of these prompts would you use next week?
  • Where did it get something about your practice or your patients obviously wrong?
  • Anything else that came up that you'd like to talk through?
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Session 1: Summary

  • A bare request gets a generic answer; context about you and your patients turns it into something you'd use
  • When you're not sure what the AI needs, ask it to interview you
  • Role, context, task, constraints, output: context and constraints are the parts most often missing
  • AI drafts from everything it has read; check what you'd act on against the standard you're accountable to
  • What goes to the patient is your decision

The habit to keep: start with context.

Break (10 minutes)

§04 · development

Session 2 ·
Professional development

  • Working with research : questioning a paper, checking what AI tells you, finding sources
  • Practice : turning a guideline into changes you can make on Monday
  • Your own learning : a plan built from what you already know, that lasts longer than one conversation
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Bring yourself into the conversation

  • Ask most tools for "a summary" and you get the same generic, surface-level summary everyone else gets
  • What makes it useful is what you bring: your profession, your patients, what you're unsure about, what you'd change
  • Every question you ask puts more of your context into the window, and the answers become more specific to you
  • That's true for everything in this session: papers, guidelines, learning plans
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Start a project for your learning goal

  • Projects (Claude), Custom GPTs (ChatGPT), and Gems (Gemini) hold files, instructions and past work around one goal, e.g. "improve my management of persistent neck pain"
  • Save what matters into the project: papers, guidelines, summaries, your learning plan
  • Start a new conversation inside the project for each task; it begins with the project's context, without the clutter of a long thread
  • Shorter threads also stretch free-tier limits further, and earlier material doesn't fade as much
  • Free tiers limit how many projects you can have; one per learning goal is enough
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Have a conversation with a paper

I've attached a paper on [acupuncture or massage for chronic neck pain]. I'm a [profession]. Before you summarise anything, let me ask you about it.

Then ask (10 minutes):

  • What did they measure, and what did they leave out?
  • Would this apply to a typical patient of mine who [is 72 / works nights / has had pain for ten years]?
  • Can you explain the statistics in plain language?
  • Where do the authors claim more than their data shows?

Then ask for the summary (later, compare it to a simple request for a summary)

§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Two ways to judge the conversation

  • We rarely ask a well-read colleague for the numbers when they say "most patients" or "works better"; we judge them by what we know of them and the topic
    • Facts : for any claim you'd act on, find the sentence in the paper yourself; watch for overstated effects, subgroups presented as the main result, missing limitations
    • Value : did it change how you think about the paper? What questions did it raise that you hadn't considered? Would you bring it up with a colleague?
  • Bring the same mix of trust and scrutiny here: check what you'd act on, and take the rest as a conversation worth having
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Ask it for references

Search PubMed and Google Scholar for five peer-reviewed papers on [shared decision-making in rehabilitation] published since 2020. For each, give the full reference, a one-line summary of the finding, and a link to the paper.

  • Naming the databases, a date range and asking for links makes every reference easy to check
  • Look each one up: does it exist, are the authors and year right, does it say what the AI claims?
  • Most tools now search the web when they judge it useful; asking it to search makes sure it does
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Putting a guideline into practice

Here is a recommendation from [the World Falls Guidelines] on [strength and balance training for older adults at risk of falling]. I'm a [profession] in a private clinic with [30-minute appointments, patients who self-refer, no group space]. Help me work out what I'd change on Monday: what I'd do differently, what gets in the way, how I'd explain it to a patient, and how I'd know it was working.

  • Use a guideline from your own field (10 minutes)
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Commission it like a scoping review

  • Sources : name them, e.g. peer-reviewed studies and named guideline bodies, a date range, the kinds of study you'll accept
  • The question : population, intervention, outcome, setting
  • The plan : Deep Research proposes a search plan first; edit it before it runs (you can approximate this without Deep Research)
  • The evidence : ask how strong it is and where the gaps are, as well as what it found (the reliability of this kind of prompt is debated)
  • The check : trace three claims back to their sources
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Plan your own learning

I want to develop my practice in [area]. Ask me questions one at a time to find out what I already know and where my gaps are. Then give me a three-month learning plan, with free sources where possible, and ten questions I could use to test myself at the end.

  • One question at a time keeps it an interview, so the plan comes from your answers
  • Push back on sources you can't access or don't trust
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Take the plan with you

  • A long-term plan outlives any single conversation, so get it out of the chat (the same applies for generated context)
  • Ask for it as a downloadable file, a checklist to paste into your notes, or a calendar of dates
  • Save it into your project, so every new conversation there starts with the plan
  • Ask for a restart prompt, i.e. a paragraph that tells a new conversation where you're up to (like a handover report):

Write a short prompt I can paste into a new conversation in a month's time, so you can pick up this plan where I left off.

§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Session 2: Summary

  • Generic questions get generic summaries; your questions make the answers specific to you
  • A project keeps one goal's files and instructions together; start each task in a new conversation inside it
  • Judge a conversation on facts (find the sentence yourself) and on value (did it change your thinking?)
  • Name the databases and dates, and ask for links, so every reference can be checked
  • Long-term work outlives a conversation: save it as a file, and keep a restart prompt

The habit to keep: bring yourself into the conversation.

Break (10 minutes)

§05 · admin

Session 3 ·
Administrative efficiency

  • Kinatex's rules : Plume for clinical notes, and what can go into other AI tools
  • Templates : reusable documents with no patient details in them
  • Everyday admin : emails, procedures, policies
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

AI can save you the most time on admin

  • Much of a clinician's week goes on writing that isn't treatment: reports, letters, emails, forms
  • Most of it follows patterns, which is what AI does well
  • Mistakes in an admin draft are easy to spot and cheap to fix, so it's a good place to build the habits from Sessions 1 and 2
  • It's also where patient information is closest to hand, so the privacy risk is highest
  • So we start with what can and can't go into an AI tool, then put AI to work within those limits
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Kinatex's rules for AI

  • Plume IA : Kinatex's approved tool for clinical notes, for OTs, physiotherapists and physiotherapy technologists
  • Clinical data : only in AI tools Kinatex has approved; never in the free ChatGPT, Claude or Gemini accounts we're using today
  • Responsibility : every note, recommendation, plan and report stays yours, whoever or whatever drafted it
  • Everything else today : general AI tools, with no patient data at all (Kinatex, Procédure encadrant le logiciel Plume)

Canada has no federal AI law yet. Canada's privacy commissioners say personal information needs consent and safeguards before it goes into a generative AI tool (Office of the Privacy Commissioner of Canada, 2023).

§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Use an invented patient

Don't
Here are my notes on a patient: [name], 54, Plateau Mont-Royal, CNESST claim, seen 12 Mar for…

Do
A typical patient of mine might be someone in their mid-fifties, a manual worker, off work after a lifting injury, with [the kind of presentation I see often]…

  • An invented patient typical of your caseload gives AI everything it needs to help you think
  • Free tools send what you type to servers outside Quebec, and some train on it by default; check your settings
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Templates, kept in a project

I regularly write [letters to referring physicians / discharge summaries]. Before you draft anything, ask me who reads it and what they need to decide. Then build a template: headings, what goes under each one, and [brackets] for every patient detail. I'll fill the brackets in myself, outside this chat.

  • e.g. a letter to a referring physician, using the Kinatex follow-up note headings: subjective, objective, analysis, plan, questions and recommendations
  • A project holds your templates, so each new document starts from the right one
  • If you keep making the same changes, describe them, ask the AI to update the template, and replace the old one in the project (10 minutes)
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Everyday admin

  • A difficult email : a patient disputes a late-cancellation fee; draft a reply that holds the policy and keeps the relationship
  • A notice in two languages : a holiday-closure notice for the waiting room and the website, in French and English
  • A clinic procedure : turn "how we do new-patient bookings" into a one-page checklist for new reception staff
  • What frequent administrative task takes the most time each week?
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

An AI policy for your patients

Create an outline for an "AI use and disclosure policy" for a private rehabilitation clinic in Quebec. Cover what AI is and isn't used for, what patient information never goes into AI tools, and how patients can ask questions. Use plain language a patient would read. Note where Quebec or Canadian privacy law applies, and flag anything I should check with a professional.

  • Edit the prompt so it describes what's true in your clinic
  • Kinatex's Plume procedure tells staff what they may do with AI; this policy tells patients, in plain language, how AI is and isn't used in their care
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Session 3: Summary

  • Clinical data goes only into tools Kinatex has approved; everything else today uses no patient data at all
  • An invented patient typical of your caseload gives AI everything it needs to help you
  • A template holds the structure; you add the patient details yourself, outside the chat
  • Admin writing follows patterns, which is where AI saves time soonest
  • Patients should be able to read, in plain language, how AI is used in their care

The habit to keep: invent the patient, so real records stay where they belong.

Break (10 minutes)

§06 · business

Session 4 ·
Business operations

  • Who you want to reach : a patient group you'd like to see more of, described as a persona
  • What you say to them : content written for that person, in your own voice, across formats
  • Where they find you : your website, and a proposal for a new service
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Most small businesses now use AI

  • 89% of small businesses use AI, up from 36% in 2023; average reported return on investment is 3.7x (U.S. Chamber of Commerce, 2026; McKinsey, 2026)
  • 18–25% average cost savings where AI is built into core workflows (McKinsey, 2025)
  • Most of those savings come from tasks like the ones you've done today

These figures come from businesses in general, outside health care. Read them as a direction of travel; they don't tell us what's true in clinics today.

§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Patient personas

  • A persona : a description of one typical person in a group you want to reach, e.g. their age, work, worries, what they search for and what stops them booking or engaging
  • Marketing teams use them so every piece of content is written for someone in particular rather than a generic cohort
  • You know some of this from your own patients; AI can fill in the gaps, and you correct what it gets wrong
  • Everything else in this session builds on the persona: the content, the voice, the website, the new service
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Name your client groups

On paper, not in the AI (5 minutes)

  1. List up to five patient groups that matter for your caseload: some you already see, some you'd like to see more of
  2. Pick one, ideally a group you don't yet see much of
  3. For that group, write down what you know: who they are, what brings them in, what might stop them booking
  • Use real groups from your own practice
  • Whatever you can't answer in step 3 goes to the AI next
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Build a persona

With the AI (10 minutes)

I'm a [profession] in [Montreal / your area]; I mostly see [groups X and Y]. I'd like to see more [group Z], and here's what I know about them: [your notes from step 3]. Build this group out as a full persona: who they are, what they'd search for at 9pm when the problem flares, what would make them book, what would stop them. Ask me questions before you start.

  • Four parts: who you are, the group, what you want back, an instruction to question you
  • Adapt the wording to your own practice
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Content for that person

I've attached a persona representing a group I'd like to see more of. Draft a blog post or a service page for our website, written for this person, that answers the question they'd be searching for. Professional but conversational, under 500 words.

  • They aren't your patients yet, so write for where they'll find you: your website, a blog post, a social media post
  • The persona's details are partly invented; edit them to match what you know, and the content improves as the persona does
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Teach it your voice

I've attached three things I've written: a patient email, a website page and a social media post. Describe how my writing comes across: its main characteristics, what it seems to be trying to achieve, and how it would make a reader feel. Then rewrite the draft so it does the same.

  • You don't need to describe your own voice; let AI describe it from your examples
  • Correct anything that doesn't sound like you, or add how you'd like a reader to feel; feed this back to the AI
  • Keep the description, and paste it into future prompts
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

One piece, five formats

Take the piece we just wrote. Turn it into: a 150-word social media post, a waiting-room handout in plain language, a three-line SMS follow-up for patients seen in the last month, a short script I could record on my phone, and one paragraph for the clinic website.

  • Change the five to whatever fits your clinic, or ask for fewer
  • Try other formats too, e.g. an infographic or a flowchart
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Your website, from another perspective

I've attached screenshots of our clinic homepage. Review it as if you were a prospective patient in pain, deciding whether to book. Tell me: what service do you think we offer, what would make you hesitate, and where do you get stuck? Then rank five changes by their likely effect on bookings.

  • Screenshots let it read the page as a page: what you see first, what competes for attention, where the booking button is
  • If your free tier limits uploads, paste the page text instead
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Pitch a new service

Act as a business consultant for a private rehabilitation clinic network. Ask me what you need to know, one question at a time. Then draft a proposal for [a new service, e.g. a return-to-running clinic, a falls-prevention group, a workplace injury programme], with the patients it serves, objectives, milestones at 3, 6 and 12 months, and what would tell us it's working.

  • On Gemini : try Deep Research for the market and evidence background, using the scoping-review principles from Session 2
  • On the others : the interview route does most of the same work
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Session 4: Summary

  • A persona turns "our patients" into one person you can write for; AI fills the gaps and you correct them
  • Let AI describe your voice from your own writing, then keep the description for future prompts
  • One piece of writing can become several formats in a single prompt
  • Screenshots let AI look at your website the way a prospective patient would
  • For a bigger idea, like a new service, let AI interview you before it drafts anything

The habit to keep: write for one person.

§07 · patients' AI

Your patients use AI too

  • Before they see you : many patients have already asked an AI about their symptoms
  • In the session : they bring what it said, and some trust it more than you might expect
  • Your part : how you respond shapes the relationship, and you can help them use it better
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI
  • 71% of psychologists say patients already discuss AI-sourced advice with them (APA, 2026)
  • 68% noticed patients feel validated or supported by it (APA, 2026)
  • 27% of people asking for personal guidance asked about health, ahead of career, relationships and money (Anthropic, 2026)
  • 230 million people ask ChatGPT health questions every week (OpenAI, 2026)
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Patients value AI guidance

  • Patients are arriving with a second opinion that didn't come from a human, and increasingly, they trust it
  • It's useful to them and imperfect, like much of the health advice they already get from friends, forums and social media
  • It's there when you aren't: at 9pm when the pain flares, with unlimited time and patience
  • Responding to this well is becoming part of clinical practice
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

When a patient's AI disagrees

A patient starts a session with: "ChatGPT told me my pain is a disc problem and I shouldn't be exercising."

  • What's your honest first instinct, and what would the patient take from it?
  • What might the AI not know about this person that you do?
  • How do you want the conversation to land, so they keep coming back to you?
  • Engaging with why you disagree, e.g. what the AI couldn't know about them, protects trust better than dismissing it (Telehealth.org, 2026)
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Turn it into a guide for your team

Draft three ways I could respond that take their research seriously while explaining what the AI couldn't know about them. Then turn the best one into a one-page guide, split by role: reception and clinician.

  • Everyone in the clinic will meet this conversation, at the front desk as well as in treatment
  • Agree the response as a team, so a patient hears the same message from whoever they speak to
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Teach patients to ask AI well

Knowing how to ask an AI about your health is part of health literacy now. You can teach it the way you teach any self-management skill:

  • Ask open questions : "What else could explain this?" gets more than "Could this be a disc?"
  • Give useful context : age, what's changed, what they've already tried
  • Share carefully : leave out names and anything else that identifies them, and check the app's privacy settings
  • Bring it back : show you what the AI said, so you can work with it together

They're the habits you practised today, passed on.

§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

What you built today

  • A patient handout and an exercise programme, checked against a guideline
  • A conversation with a paper, a guideline turned into practice, and a learning plan you can take with you
  • A reusable template with no patient details in it, and an AI-use policy outline for patients
  • A persona, content in your own voice, and a proposal for a new service
  • A way to respond when a patient's AI disagrees with you
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

A handover report

Create a handover report from this conversation. Give me 10 principles I should take away and use in my practice.

  • Run it at the end of any conversation worth keeping
  • Pair it with a restart prompt, and you can pick up any piece of work later
§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

Read, review, correct, approve

Kinatex's rule for Plume works for any AI tool:

  • Read it in full
  • Review it critically
  • Correct it where it needs correcting
  • Approve it yourself, before it goes into a record or to anyone else

Whoever or whatever drafted it, the note, the recommendation or the report is yours (Kinatex, Procédure encadrant le logiciel Plume).

§01 why it matters §02 how AI works §03 patient education §04 development §05 admin §06 business §07 patients' AI

One thing to try this week

  • Pick one task from today that you do every week
  • Do it once with AI, using context, an interview and a check
  • Keep the prompt that worked, and use it again next time
questions

Thank you

mrowe@lincoln.ac.uk

QR code linking to the page for this masterclass, with the slides and handout
slides and handout

kinatex education committee masterclass · montreal (online)   2026-10-24