Physical Therapy Policies and Procedures: A Clinic Manual Drafted With AI

Write down what your clinic really does at each step (intake, consent, cancellations, records, complaints), give those notes to an AI tool and ask for a draft policy in a fixed layout. Then a named clinician checks every section against current professional guidance and signs it off in a log. AI drafts; people carry the responsibility.

The common failure is a manual that reads well and describes a clinic that does not exist. AI is very good at confident, tidy policy language, and very bad at knowing whether your front desk follows it. So the method below starts from your routines, not from a blank prompt, and ends with a sign-off that records who checked what.

Follow me on Instagram@sagnikteaches

What a small clinic's manual needs, and what it can leave for later

A manual of 80 pages nobody opens helps no one. For a clinic of three to ten staff, a working manual is a dozen short policies, each one to two pages, in the same layout. Start with the ones that cause real trouble when they are missing.

Connect on LinkedInSagnik Bhattacharya

Add AI use as its own short policy once you use it anywhere near patient information. Later additions such as safeguarding, lone working, chaperones, social media and staff supervision can follow. The aim is to cover the first eight well rather than twenty badly.

Subscribe on YouTube@sagnikteaches

Gathering your real routines first

The raw material is how your clinic works today. The quickest way to capture it is to talk. Ask the person who does the task to describe it aloud, record it on a phone and transcribe it, or type quick bullet notes. Ten minutes per policy is enough. A transcript is messy, and that is fine, because the AI tidies it. What it cannot do is invent facts you did not give it.

Capture five things for each policy: the trigger (what starts it), the steps in order, who does each step, what gets recorded and where, and what happens when it goes wrong. Here is an illustrative set of rough notes for cancellations, as the front-desk lead might dictate them.

Cancellations. We ask for 24 hours notice. If someone cancels with
less than that, or doesn't turn up, we send a message the same day.
First time we let it go and just remind them. Second time in a plan
of care we tell them about the late fee, which is half the session
price. We waive it if they were ill, we note why. Therapists don't
decide fees, the front desk does, but the lead can override. We try
to fill the slot from the waiting list. After two no-shows in a row
the lead therapist calls them to see if the plan still suits.

The notes are rough, but every decision is in there. Notice the things to check later: is the fee lawful and clearly agreed in advance, is a waiver for illness recorded consistently, and does the patient know about the policy before they are charged? Those are not for the model to settle.

From dictated notes to a finished cancellation policy

Give the AI the notes, a fixed layout and a few rules. A layout that every policy shares makes the manual easier to read and to audit. This one works for most clinics.

You are drafting a policy for a small physical therapy clinic's
manual. Use only the facts in my notes below. If something needed is
missing, write "CLINIC TO CONFIRM:" followed by the question,
rather than inventing an answer. Do not cite laws or standards.

Layout, with these exact headings:
Purpose (one sentence)
Who this applies to
What patients are told, and when
Steps (numbered, with who does each)
What is recorded and where
Exceptions and who can approve them
Review date, owner and version (leave blank for me to fill in)

Plain English, short sentences, no jargon. Under 350 words.

My notes:
[paste notes]

An illustrative extract of what comes back:

Purpose
To set out the notice we ask for when an appointment is cancelled,
and what happens when notice is not given.

Steps
1. Front desk records the cancellation, the time it was received and
   the reason, if given.
2. If notice is under 24 hours or the patient did not attend, the
   front desk sends a message the same day.
3. First late cancellation in a plan of care: reminder of the policy,
   no fee.
4. Second late cancellation: front desk explains the late fee (half
   the session price), unless an exception applies.
5. CLINIC TO CONFIRM: Where in the patient's record is the reason
   for the waiver noted?
...

This is a good draft precisely because it asked a question rather than guessing. The owner answers it, sees that the fee must also appear on the booking confirmation and the website, and adds that. The clinician then checks with the insurer or adviser whether the fee wording is acceptable where the clinic operates. Only then does the policy go in the manual.

Consent: where drafting needs a clinician in the room

Consent is the section where an AI draft can be fluent and wrong. What counts as valid consent, how it must be recorded and what applies to minors or patients who lack capacity are set by law and professional standards that differ between places and change. Use AI for the structure and plain wording, and let the lead clinician check every sentence against current standards.

A helpful split is to have the model draft the process and leave the rules to you. The process: who explains what, at which visit, in what form, how a refusal or a withdrawal is recorded and who is told. The rules: what must be explained, to whom and when, which you take from your professional body's guidance. In the draft, mark each rule with "CLINIC TO CONFIRM" until a clinician has supplied it from the source, and note the source name and date next to it. If your clinic uses AI note-taking, the consent wording in telling patients you use AI note-taking is a ready starting point to adapt.

Records, privacy and how long things are kept

The records policy covers what is written, how quickly, who may read it and how long it is kept. Two of those are yours to describe (the timing and the access in your clinic) and two are for you to verify (the content requirements and the retention period). Retention periods are set by regulators and sometimes insurers, and they vary by place and by patient type. Do not accept a number the model offers. Find the rule from your own regulator or insurer, write it into the policy with its source and date, and have the owner confirm it.

A short privacy and AI paragraph belongs here too. State plainly which tools may touch patient information, which may not and how staff check before pasting anything into a chat. The checklist in patient data and AI for small practices can be copied nearly as it stands. If you use an AI scribe, name it in the policy and link to its vendor's data terms; the comparison of AI scribes for physical therapy helps you choose what to name.

Complaints: a procedure people can follow when they are upset

A complaints policy is read under pressure, by someone with an angry patient in front of them. It needs plain steps and named people. Draft it from how complaints are really handled, and add timescales you can keep. Here is an illustrative skeleton, filled in for a four-clinician clinic.

1. Receive. Any staff member who hears a complaint thanks the patient, writes down what they said in the patient's own words and tells them who will respond and by when.
2. Log. The practice manager records the complaint in the complaints log the same day, with a reference number, date and the person assigned.
3. Acknowledge. The assigned person contacts the patient within two working days to confirm it has been received.
4. Look into it. The assigned person speaks to those involved and reads the relevant record. A clinical complaint goes to the lead clinician.
5. Reply. A written reply within ten working days, or a message saying why more time is needed and when to expect a full answer.
6. Learn. The owner reviews the log every quarter and notes any changes made.

The two-day and ten-day figures above are illustrative. Choose timescales you can keep and, if your regulator or insurer sets any, use theirs. AI can polish the wording and draft template letters for acknowledgement and reply, but a complaint file never goes into a chat. If you want a model to help with a reply, write a version with placeholders and no case details.

The sign-off log that makes the manual trustworthy

This log is the part most clinics skip and the part that matters most later. It records who checked each policy, what against and when. Keep it as a one-page table at the front of the manual, with one row per policy and per version.

The rows above are illustrative. The columns that matter are "checked by" and "checked against": a policy checked against nothing has not been checked. If a regulator, insurer or patient later asks how your procedures were created, the log shows a considered process rather than a pasted draft.

Roles, not names: what stays out of the drafting chat

A policy describes a process and has no need for anyone's name. Use roles ("front desk", "lead clinician") instead of staff names, and never paste patient cases, complaint details or staff performance issues into a chat. If a real incident prompted the policy, describe the lesson in general terms. Use a business plan whose terms say content is not used for model training (ChatGPT Business and Claude Team both say so on their pricing pages, checked November 2026), and keep the drafting inside a Project or workspace so the prompt layout and your boilerplate stay in one place.

Intake, a website clash and an emergency drill

Intake: turning a form into a procedure

A clinic has an intake form but no written steps for what happens with it. The practice manager dictates: form emailed on booking, checked for completeness at the front desk the day before, red-flag questions reviewed by the clinician before the patient is seen, identification checked at arrival. The AI turns that into numbered steps with owners. The clinician then checks the red-flag wording against professional guidance and adds the actions to take on a positive answer.

A policy that clashes with the website

While drafting the cancellation policy, the owner finds the website says 48 hours' notice and the booking emails say 24. The AI cannot see that; the owner does, because the draft asked "CLINIC TO CONFIRM: where is this notice period stated to patients?". Fixing the conflict before publishing the manual prevents a patient dispute later. Treat every such question as a free audit.

An emergency procedure tested in a drill

The emergency policy lists who calls for help, who stays with the patient, where equipment is and how the incident is recorded. A ten-minute drill at a staff meeting shows that nobody knows where the first-aid kit is kept. The policy and the room both change. AI helped draft the procedure; people tested it and found the gap.

A short AI-use policy the whole team can follow

Once any member of staff uses an AI tool for clinic work, the manual needs a page saying how. Draft it from three questions put to the team: which tools do you use today, for what, and what do you paste into them? The answers are often a surprise. A receptionist may be using a free personal chatbot to tidy emails; a clinician may be dictating into a phone app. Better to know now. A workable one-page policy has five lines.

1. Approved tools. A named list (for example, the clinic's business-plan chat workspace and the practice software's own features). Anything else is not for clinic work.
2. Allowed uses. Drafting templates, tidying non-patient emails, summarising public guidance, drafting policies from routines.
3. Never pasted. Patient names, identifiers, notes, complaint details, staff personal information.
4. Always checked. Anything AI drafts that a patient or payer will see is read by a named person before it goes.
5. Reporting. Anyone who pastes something they should not tells the practice manager the same day, with no blame, and the incident is logged.

That last line matters most. A team that fears punishment hides mistakes; a team that can report one quickly lets you delete a chat and learn from it. The template in an AI incident response plan for small businesses gives the steps for what happens next.

Turning a policy into a one-page card for the front desk

A policy is a reference document; a card is what people use at the desk. When a policy is signed off, ask the model for a condensed version: the same steps, in a half-page, with the exact phrases staff can say to patients. For cancellations, an illustrative card says: what to say when a patient calls within 24 hours, what to say to a first-time late cancellation, where to record the reason and who to ask for an exception. Print it and put it by the phone. The model drafts the card; the owner checks it says nothing the full policy does not.

The same approach works for the equipment and cleaning policy, where a card becomes a weekly checklist: treatment couches wiped and checked between patients, exercise equipment inspected weekly, anything damaged tagged and removed, a named person who signs the sheet each Friday. A checklist like this is easy for AI to draft from notes and easy for a clinician to verify, which makes it a good second policy to write after cancellations.

One last practical point on versions. Keep the current manual as a single file with the version and date in the filename, and move superseded copies to an archive folder rather than deleting them. When a policy changes, the log gets a new row, the old row stays, and the briefing note says in two lines what changed and why. An archive lets you answer "what was our policy on this date?", which is the question that arises in a dispute, and it costs nothing to keep.

Why staff stop trusting a policy manual

• It describes an ideal clinic. A policy says all notes are written within an hour; in practice they are done in one batch after the last patient. Staff learn to ignore the manual. Fix: draft from actual routines and rewrite the policy or the routine, whichever is wrong.
• Invented standards creep in. The model cites a retention period or a standard that sounds right. Fix: instruct it to write "CLINIC TO CONFIRM" rather than cite, and verify each external requirement at the source.
• The manual is never reviewed. A date passes. Fix: review dates in the log and a calendar reminder for the owner.
• Staff were never told. The manual exists but nobody read it. Fix: a short briefing for each new or changed policy, and a signed acknowledgement from each team member. Draft the briefing notes with AI if it saves time.

A drafting timetable that fits around clinic days

Eight policies in one go is a project nobody finishes. A steadier plan is one policy a week over two months. Each week: ten minutes of dictated notes on Monday, an AI draft on Tuesday, the owner's gaps filled on Wednesday, a clinician's check on Thursday and a staff briefing at the Friday huddle. That is roughly an hour of total effort per policy, spread across people. After eight weeks you have a manual, a sign-off log and a team that has seen every policy once.

Then test it. In week nine, give a part-time or newer team member three situations and ask them to answer from the manual and the desk cards alone: a patient cancelling with two hours' notice for the second time, a patient complaining at the desk about a bill, and a patient asking for a copy of their notes. Watch where they hesitate, guess or ask a colleague. Each of those moments marks a policy with a missing step, an unclear owner or a card that says less than the policy. Fix those first, record the change in the sign-off log, and repeat the test with a different person after the next review.

If you want more on the method, the general guide to writing SOPs with AI covers the notes-to-procedure step in detail, and writing an employee handbook with AI handles the staff-facing companion.

Questions before you draft the manual

Can AI write the whole manual for me?

It can draft the structure and the wording, but only from your real routines. A manual that describes what you do not do is worse than none. Give it your actual steps, then have a qualified clinician and your adviser check every section before anyone relies on it.

Who should sign off the manual?

A named senior clinician for clinical content, the practice owner for operations, and where relevant your insurer or legal adviser for consent, records and complaints. Record each sign-off with a name, a date and a version number in a log kept with the manual.

How often should policies be reviewed?

At least once a year, and whenever a rule, a payer requirement or your own process changes. Put the review date on the front of each policy and a recurring reminder in the owner's calendar so the review actually happens.

What should never go into an AI chat when drafting?

Patient names, case details, staff personal information and anything from a complaint file. Describe routines in general terms, use placeholders and keep real cases out. A manual describes how you work, not who you treated.

Is a template from a professional body better than an AI draft?

A professional body's guidance is the authority on standards, so read it and cite it in your manual. AI is useful for turning that guidance plus your own routines into plain, clinic-specific procedures that staff will actually read.

Further reads

• How HR Consultants Use AI for Policies, Letters and Cases — How HR specialists approach policies and letters with AI.
• Dance Studio Handbook: Parent Policies Drafted With AI — Another small business turning routines into written policies.
• Physical Therapy Prior Authorisation: Track Visits With AI — A procedure for authorisations you can add to the manual.
• AI Receptionist for Physiotherapy Clinics: Cost and Booking Setup — Draft front-desk scripts that match your cancellation policy.
• Is an AI Scribe Worth It for a Small Private Clinic? — Decide whether an AI scribe belongs in your records policy.
• How to Set Up Human Review for AI Work Without Slowing Down — The reviewer habit behind a sign-off log, applied to other AI work.
• Best AI to Help With PT Documentation: Avoiding Cloned Notes — Use AI for PT documentation without cloned notes: what auditors look for, a per-visit detail checklist, prompts and a therapist review step.
• Physio Discharge Letter: Template and AI Draft for GP and Patient — For practice owners: an approved discharge letter template, how AI drafts the GP and patient versions from a clinician summary, and the sign-off rules.
• Using ChatGPT for Physical Therapy: Referring-Doctor Updates and Outreach — Build a library of referral update and outreach templates in ChatGPT, then fill the patient details from your own records, never the chatbot.
• AI Use in Physical Therapy: Which Uses Need Clinician Sign-Off — A clinic owner's ladder for AI in physical therapy: which uses run alone, which need a clinician to sign, and which stay human.
• What Policies Should a GP Practice Have? Checklist and AI Drafts — A GP practice policy checklist grouped by risk, with a review calendar and tested prompts for drafting each policy with AI before a partner signs it off.
• Will Payers Accept AI-Written Physical Therapy Documentation? — Payers judge the note, not the tool: what reviewers look for, and the review, signature and audit-trail routine that makes AI drafts safe.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: ChatGPT Business pricing page (openai.com) and Claude pricing page (claude.com), both checked 4 November 2026, for no training on business content by default. Professional body guidance, regulator requirements and insurer conditions differ and change, so every clinical and legal point in a draft must be checked against your own current sources. Examples are illustrative.