AI Medical Scribe for Clinicians: Shared Templates in a Group Practice

A group practice standardises AI scribe templates by sharing the structure and leaving the phrasing personal. Lock the headings, required fields and safety prompts in one core library owned by a named clinical lead, then let each clinician keep a personal copy that inherits the core and adds their own style instructions.

Most practices get this backwards. They either let every clinician build their own template, which produces a dozen incompatible note shapes the moment someone covers a colleague's list, or they impose one rigid template and watch half the doctors stop using it. The split below avoids both.

Follow me on Instagram@sagnikteaches

What belongs in the shared core and what stays personal

A scribe template is two things at once: a structure (the headings and fields a note must have) and a set of instructions about how to write inside it (tone, abbreviations, how much detail). Treat them separately. Structure is a practice decision because it affects handovers, audits, coding and anyone reading a note they did not write. Style is a clinician's decision because it affects how quickly they can check and sign the note.

Connect on LinkedInSagnik Bhattacharya

Vendors build this inheritance in different ways, so check what your product actually offers before you design around it. As of November 2026, Heidi's pricing page shows three plans: Free, Clinician and a custom-priced Teams plan. Clinician includes custom templates for one person, while team template sharing, template configuration and team management controls appear only on Teams, which lists shared templates, sessions and patient information. Freed puts note sharing, admin dashboards and custom configurations on its custom-priced Groups plan, and its plan comparison lists a team template library and custom note-retention settings among the team and admin features. Neither page says how a personal copy inherits changes to a shared template, so test that in a trial (see below) and confirm the tier with each vendor before you budget.

Subscribe on YouTube@sagnikteaches

Four things to test in a vendor trial

1. Locking. Can the owner stop other clinicians deleting or renaming a heading in a shared template? Try it from a clinician's login, not the admin's.
2. Propagation. When the owner edits the shared core, does a clinician's personal copy pick up the change, keep the old version, or ask? Each answer needs a different change-control routine.
3. Visibility. Can an admin see which template and version produced a given note? Without this, the audit below has to rely on clinicians remembering.
4. Export. Can you export the template text? If you leave the vendor, the library should leave with you.

Write the four answers down for each product you trial. A scribe that fails on locking can still work, but you will run more of the control through process, as the next sections describe.

An illustrative practice: six clinicians, three starting styles

Say a practice has four doctors, one nurse practitioner and one physiotherapist. Before standardising, the practice manager asks each of them to paste three recent notes, with patient details removed, into a folder. Reading eighteen notes takes about an hour and shows three habits: two doctors write long narrative histories, two write terse problem lists, and the nurse practitioner and physio write in a structured SOAP shape.

The temptation is to pick the best-looking style and impose it. Don't. Instead the practice lead writes a core with these headings: Reason for visit, Relevant history, Examination, Assessment, Plan, Safety-netting, Follow-up. The narrative writers keep a long Relevant history; the terse writers keep a short one. Both are valid because the heading is fixed and the length is not.

The cost of this exercise is mostly time: roughly an hour for the review, two hours for the clinical lead to draft the core, and thirty minutes per clinician to set up a personal copy and test it on one real appointment. Call it six hours across the practice, one time, plus the pilot week. That is an illustrative estimate, not a vendor figure.

Build the library in three layers

1. Core templates. Owned by the clinical lead. Locked for everyone else. Four to six at the start: new patient, standard follow-up, procedure, telehealth, and one for each service line that genuinely writes differently.
2. Service-line templates. Inherit the core and add only the fields that line needs, such as a pain score for musculoskeletal visits or a growth-chart line for children.
3. Personal copies. Each clinician duplicates a core or service-line template and adds style instructions. They can reorder nothing and delete no required field.

If your scribe cannot lock fields, the control moves into process: the library lives in one shared document, clinicians paste updates from it quarterly, and the audit (below) catches drift. That is clumsier but workable.

A core template, filled in

This is the instruction block for a standard follow-up. It is a sample for you to adapt, not vendor wording.

Write a follow-up consultation note using exactly these headings, in this order:
Reason for visit; Relevant history; Examination; Assessment; Plan; Safety-netting; Follow-up.

Rules:
- Only include what was said or done in the consultation. Do not infer.
- If allergy status was not discussed, write "Allergies: not discussed" under Relevant history.
- If no safety-netting advice was given, write "Safety-netting: none recorded".
- Record medicines with dose and frequency exactly as stated.
- Put any numbers (blood pressure, weight, results) in the Examination section only.
- British spelling. No bold text, no emojis.
- Mark anything unclear with [CHECK] so the clinician sees it before signing.

The two "write this exact phrase" rules matter most. They turn silence into a visible gap. A scribe left to fill a missing safety-netting line will often produce something plausible; an instruction to write "none recorded" gives the signing clinician a prompt instead of a fabrication.

Personal style without breaking the structure

The personal layer is a short block appended to the core. Two real-looking examples:

Clinician A (long narrative): Write Relevant history as two to four sentences of prose, chronological. Spell out abbreviations on first use.

Clinician B (terse): Write Relevant history as a bulleted list, maximum five bullets, standard abbreviations (HTN, T2DM, SOB) allowed.

Both produce notes with the same seven headings. A third clinician reading either can find the plan in two seconds. That is the whole point: the standardisation is invisible to patients and almost invisible to the clinicians, but the audit and the handover both get easier.

A before and after shows how it reads. Before, a covering doctor opens a colleague's note and has to scroll through three paragraphs of prose to learn the patient is waiting for blood results. After, the Plan heading says "Awaiting bloods; review in 1 week" in the same spot every time, and the long narrative is still there above it for anyone who wants the story.

The sampling audit that keeps the library honest

Standard templates drift. A clinician adds a line, a vendor update changes how a heading is phrased, someone copies an old version. A light audit catches this without becoming a second job.

1. Sample. Each month, pick five notes per clinician at random from the previous four weeks. For a six-clinician practice that is thirty notes.
2. Score. One page, five yes/no questions: all headings present and in order; required fields present or explicitly flagged; plan matches what the clinician would say they did; no invented content; no content from another patient.
3. Time. About three minutes a note, so ninety minutes a month for the reviewer. Rotate the reviewer so it is not always the clinical lead marking their own team.
4. Act. One failure is a conversation. Three of the same failure across clinicians means the template is wrong, not the people.

The fourth and fifth checks matter most for patient safety. Invented content is the known failure of generated notes; content from the wrong patient is the failure of a busy clinic with two windows open. If either turns up, treat it as an incident under your usual process, not as an audit note.

A filled-in audit result

Illustrative: in the first month the thirty-note sample shows 27 notes with all headings and three missing Safety-netting. All three are telehealth consultations. The cause is not careless clinicians: the telehealth template was copied from the follow-up one before the safety-netting line was added to the core, and nobody updated the copy. One fix, one sentence in the telehealth template, and the next month's sample is clean.

Consent wording and opt-outs that apply across the practice

Consent is the one thing you do not let each clinician improvise. A patient who sees two clinicians in a fortnight should hear the same explanation and have the same choice. One national health service's published guidance on ambient scribing products asks settings to be transparent about how information is used, to update privacy notices and explain the use before processing starts, to give patients the chance to object, and to train practitioners in gaining patients' permission. Wherever your practice sits, that is a sensible outline.

A short script for the clinician to say, and for the front desk to mirror in the waiting room, might read:

"I use an AI note-taking tool that listens to our conversation and drafts my notes. I read and correct the notes before they go on your record, and the recording is not kept after the note is done. If you would rather I didn't use it, just say so and I'll write the notes myself, and it won't affect your care."

Check that last promise about recordings against your vendor before you say it. Freed's FAQ says recordings are saved temporarily until summaries and quality checks are complete and then automatically deleted. Heidi's pricing page lists "flexible data retention" in its plan comparison without saying what the options are, so you need the detail in writing. Never put a promise in a script that you have not confirmed in the vendor's terms. The vendor security questions tutorial lists what to ask.

Making the opt-out easy to use

• Put a marker on the appointment list showing "scribe off" for patients who opted out last time, so they are not asked to refuse again.
• Record the choice in the patient's record, not only in the clinician's memory.
• Allow a clinician-level opt-out too. A doctor who dislikes the tool should not be shamed into it.
• Add a notice in reception and on the website. If consent scripts differ between sites, the one written down wins.

If you are introducing a scribe to patients for the first time, the sequencing in how to roll out an AI scribe without losing patient trust pairs well with this one.

Versioning and change control for a living library

A library with no history is a library that nobody trusts. Keep it simple: each core template has a version number, a date, a one-line reason for the last change and an owner. Store it where the practice keeps other controlled documents. When the clinical lead changes a template, they tell clinicians in the Monday huddle and note the effective date.

Vendor updates are a second source of change. When the scribe product changes its note engine or adds a feature, run one test appointment per core template before the whole practice uses the update. Spend ten minutes comparing the output with the previous version, and only then circulate. If your vendor does not announce changes ahead of time, ask for a changelog as a condition of the plan.

Five failures and how they show up

1. Template sprawl. Twenty templates because every clinician wanted a small tweak. You spot it in the template list: several with near-identical names. Fix it with a quarterly cull of anything unused.
2. The silent edit. Someone changes a core template and nobody tells the others. You spot it in the audit when a heading disappears. Fix it by limiting edit rights to the owner.
3. Copy-paste between patients. A clinician reuses the previous note as a base. You spot it when two notes share an unusual sentence. Fix it with the "content from another patient" audit question and a standing rule against reusing a note as a starter.
4. Over-trust. A clinician signs without reading. You spot it when the [CHECK] markers survive into the signed note. Search for them weekly.
5. Consent drift. Clinicians stop saying the script because it feels long. You spot it by asking three patients at the front desk what they were told. Shorten the script rather than abandoning it.

Seat costs and when to move to a group plan

Individual plans are cheap to start but make shared control impossible. Moving to a team or group plan is usually the moment the practice gets admin, centralised billing and shared libraries. Look at what the plan page for your chosen vendor lists, then ask for a quote. Freed's Groups plan and Heidi's Teams plan are both priced through sales. Heidi's individual Clinician seat was listed at $110 per user a month, billed annually, when I checked its pricing page in November 2026, which gives you one reference point to hold a team quote against. Do not budget from a screenshot you saw in a forum.

A rough way to compare: take the number of clinicians, the minutes each saves per day, and a valued hour. If six clinicians each save twenty minutes a day over twenty working days, that is forty hours a month before audit time, against about an hour and a half of audit. The sum only means something once you replace those figures with your own, measured over a pilot rather than assumed. The comparison of scribe costs for small clinics gives ranges to start from.

A four-week rollout for the standardised library

1. Week 1. Collect sample notes, draft the core, name the owner. Choose two clinicians who are willing to test.
2. Week 2. The two testers use the core for a full week. Record every edit they make to generated notes and why. Adjust the core once, at the end of the week.
3. Week 3. Roll out to the other clinicians, each with a personal copy. Display the consent notice. Run the first mini-audit of ten notes.
4. Week 4. Run the full thirty-note audit, fix what it finds, and set the quarterly review date.

If you have not yet chosen a scribe, work through the decision rule by country and record system first. Standardising on the wrong product is an expensive way to find out. If you would rather map the template library with someone, that is what my AI implementation consultation covers.

Locums, trainees and part-timers

The people who feel a shared library most are those who work across several clinicians' lists. A locum who arrives on Monday should not have to learn six personal styles. Give visiting clinicians a seat on the core templates only, with no personal overrides for their first two weeks, and a one-page card that shows the headings and the three phrases the scribe writes when something is missing. A trainee gets the same, plus a rule that a supervising clinician reads the first ten signed notes.

Part-timers raise a different problem: they forget settings. If a nurse works Tuesdays and Thursdays only, a quarterly change announced at a Monday huddle never reaches her. Put the change note into the template itself, as a version line at the foot of the instruction block, and into whatever message channel everyone reads. Redundancy is cheap here.

A sample output, and what you would fix

Illustrative output from the core follow-up template, for a made-up patient with high blood pressure:

Reason for visit: Blood pressure review.
Relevant history: On amlodipine 5 mg once daily. Home readings average 148/92. Allergies: not discussed.
Examination: BP 146/90, pulse 72 regular.
Assessment: Hypertension, above target on current dose.
Plan: Increase amlodipine to 10 mg once daily. Repeat bloods [CHECK].
Safety-netting: none recorded.
Follow-up: Review in four weeks.

Three things for the clinician to fix before signing. "Allergies: not discussed" is correct output, and the clinician should now ask or note the status from the record. "Repeat bloods [CHECK]" means the scribe heard something unclear; the clinician replaces it with the actual test and date. "Safety-netting: none recorded" is a prompt to add what was said, or to say it. Without the rules in the core template, a scribe might have written "Patient advised to return if symptoms worsen", which sounds reasonable and may never have been said.

This is also how you train new users quickly: show them one marked-up note rather than twenty minutes of slides. The three flags tell them what the tool admits it does not know, which is the part people are most likely to skip.

Keeping clinicians in the loop

A library imposed from above will be ignored; one that clinicians help shape will mostly be used. Invite each clinician to propose one change per quarter, with a reason. Accept those that improve notes for more than one person and politely decline those that serve a single habit. Share the audit results as a practice-level picture, never as a ranking of clinicians, so people keep reporting their own errors.

Finally, remember who owns the record. The scribe drafts; the clinician reviews, edits and signs. Nothing in the shared library changes that, and every template should say so in its first line so the habit is written into the tool.

Questions practice managers ask about shared scribe templates

How many shared templates does a group practice need?

Start with four to six: new patient, standard follow-up, procedure or minor surgery, telehealth, and one for each service line that writes differently. More than ten usually means clinicians are forking the library to avoid agreeing wording. Retire any template nobody used in a quarter.

Do shared templates stop a clinician writing in their own voice?

Not if you share the structure and leave the phrasing alone. Lock headings, required fields and safety prompts. Let each clinician keep a personal version that inherits those, plus their own style instructions for tone, abbreviations and level of detail.

Who should own the template library?

One named clinical lead owns it, with the practice manager handling access and version records. Changes need a short written reason and a date. Ownership split across everyone means nobody notices when a template drifts.

Can one clinician opt out of the scribe entirely?

Yes, and the practice should say so in writing before rollout. Opting out should cost the clinician nothing in standing, and the front desk needs a way to flag it on the day's list. A scribe pushed on an unwilling clinician produces poor notes and resentment.

Further reads

• Is an AI Scribe Worth It for a Small Private Clinic? — Whether a scribe pays for itself at your size.
• What Is an AI Scribe and How Does It Work in a Consultation? — The plain-English mechanism if a colleague is new to scribes.
• AI Scribe for Psychologists: Notes, Reports and Consent Checks — Consent wrinkles for talking therapies sharing your premises.
• How to Write an AI Usage Policy for Your Small Business — The staff policy your scribe rules should sit inside.
• AI Scribe for GPs: Costs, Free Tiers and the NHS Pilot Limit — Free tiers and paid plans for a surgery comparing scribes.
• Twofold AI Medical Scribe for Clinicians: Trial Checklist and Cost — Twofold's Free, Personal and Group plans as published, what the page says about audio and data, and a trial checklist to run before a small clinic pays.
• AI Scribe for Audiology: Fitting Notes, Counselling and Consent — Four AI scribes compared for fitting appointments, with filled-in fitting and counselling notes and consent wording for patients who struggle to hear.
• AI Scribe for Chiropractors: Chiro Tools vs Heidi and Jane — Chiro-specific scribes against Heidi and Jane for SOAP notes, with prices, a 90-second review routine and the payer documentation points a scribe can miss.
• AI Scribe for Optometrists: Exam Notes, Templates and a Test Plan — AI scribes for eye-exam notes compared by how they reach your EHR, with a filled exam template, consent wording and a two-week test plan.
• OpenAI for Medical Providers: Which Product Fits a Small Practice? — Which OpenAI products will sign a BAA, which will not, and how a solo clinician, small practice or developer should choose between them.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: Heidi Health and Freed pricing pages, including Freed's FAQ on audio recordings (fetched 4 November 2026); published national health-service guidance on AI-enabled ambient scribing products (fetched 4 November 2026).