Physicians in small private practice mostly pay for seven tools: two ambient scribes that draft the note from the consultation (Heidi, Freed), ChatGPT for Clinicians for cited answers and letters, Tebra AI Note Assist for notes inside the EHR, an AI phone assistant, a pre-submission claim checker and a business chat plan for admin.
Not all seven earn their cost. Two or three do, for most doctors. The list below gives what each tool does, the listed price where one is published, an example of use, and the watch-out that matters. Prices are list prices checked in November 2026 and change often, so confirm them on the vendor's page before deciding.
1. Heidi: the scribe with a free tier
Heidi listens to a consultation, with the patient's consent, and drafts the note. Its pricing page lists four plans: Free, with unlimited scribing and standard templates; Clinician at $110 per user a month billed annually, or $150 billed monthly; Practice at $180 per user a month billed annually, for practices that want shared templates, team controls, one invoice and free seats for assistants; and a custom Enterprise plan. Clinician adds advanced templates, evidence answers that take the patient's context into account during the visit, full-text journal access and a monthly allowance for Heidi's agents and routines (its automation features).
Example. An illustrative primary-care visit: the physician speaks the assessment and plan aloud near the end, and Heidi drafts a SOAP note. The physician spots that the plan includes a follow-up that was only discussed, edits it, and signs. A custom template means the note always ends with the physician's usual "advised when to seek help" line, but only because the physician put that line in the template.
Watch-out. The free plan gives only limited use of the advanced templates and a smaller weekly allowance for agents, so a busy week can hit the ceiling. Ask in writing which plan includes the agreement your privacy rules require, and how long audio and transcripts are kept. Heidi's pricing page lists flexible data retention and says Heidi does not train models on patient data; get both confirmed in the contract, not from a review site.
The Complete Excel Course with AI Integration
From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.
By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner
Buy now2. Freed: tiered by note volume
Freed prices by what you need. Its pricing page lists Starter at $39 a month (up to 40 notes a month), Core at $79 (unlimited notes, a template builder and clinical evidence answers), Premier at $104 a month billed annually or $119 monthly (adds visit summaries, patient instructions and letters, EHR push, dictation, and ICD-10 and CPT code suggestions) and a custom Groups plan. There is a 7-day free trial with no credit card.
Example. A solo physician who sees 15 patients a day needs about 300 notes a month, which makes Starter pointless and Core the sensible floor. If she writes a lot of referral letters, Premier's instant letters may pay for the $25 difference (on annual billing) by themselves: ten letters a week at five minutes each is nearly an hour of typing.
Watch-out. Freed's page says audio is stored temporarily in a secure way until summaries and quality checks finish, and then deleted. That is a vendor statement; ask for the retention terms in the contract. EHR push works into browser-based EHRs: test it on your own system with a dummy patient first.
Heidi or Freed? Try both for a week each, on your own consultations, and compare how many notes need major edits. Differences between them are smaller than the difference between how doctors talk. For help deciding between ambient capture and dictation, see AI dictation software for doctors: dictation or ambient notes.
3. ChatGPT for Clinicians: a free assistant for verified clinicians
OpenAI's help centre describes ChatGPT for Clinicians as free at launch for verified clinicians in the United States: physicians, nurse practitioners, physician assistants and pharmacists, who sign up with a ChatGPT account and pass a third-party licence check. It adds clinical search with citations, deep research across medical literature, continuing-education credits on eligible questions, and drafting help for referral letters, prior-authorisation letters and patient instructions. Individual clinicians can review and sign a business associate agreement under Settings > Agreements, and OpenAI says content shared with it is not used to train its models. A practice that wants one agreement across several users, or central admin controls, is pointed to a separate product, ChatGPT for Healthcare. These are launch terms, so check eligibility and the agreement on OpenAI's own page before relying on them.
Example. Before a complex visit, a physician pastes an anonymised clinical question and asks for a summary of the options with sources, then checks the sources. After the visit, she asks it to draft a plain-language patient information sheet for a common condition and edits it. It does not replace the scribe, because it is not built to capture the consultation as it happens, but at $0 it earns a trial.
Watch-out. It will occasionally state something wrongly with great confidence. The rule for any clinical assistant: it proposes, you verify against a source you trust. An individual agreement covers you, not your staff, so do not let the front desk use your account.
4. Tebra AI Note Assist: the note tool inside the EHR
If your practice already runs on Tebra, its AI Note Assist turns patient conversations into notes inside the EHR, where you review, edit and approve them. Tebra's page does not publish prices for its AI tools; you ask for a quote. A customer survey published by Tebra reports providers saving about nine minutes per note, which is self-reported by the vendor's customers, so treat it as a claim to test, not a promise.
Example. A three-doctor practice on Tebra asks for a quote and a trial on one doctor's list. The deciding point is not the note quality but the workflow: the note lands in the right chart field without copy-paste, and the doctor never leaves the screen.
Watch-out. The convenience is real but so is the lock-in. If you later leave the EHR, you leave the scribe. Compare the quote against a standalone scribe such as Heidi or Freed at the same number of doctors. For a scribe decision by EHR and market, see AI scribe for medical practices: pick by country and EHR.
5. An AI phone assistant for the front desk
A phone assistant answers when the lines are busy or the office is closed, takes messages and handles simple requests such as opening hours. Pricing varies. One published example is Smith.ai's AI receptionist, listed as free for 25 calls a month, then $3 a call, with a Pro plan from $150 a month for 75 calls. A healthcare-specific one is Freed's Front Desk, an AI medical receptionist in beta, listed at $149 a month per block of 250 calls for single-site practices, with warm transfer to staff during business hours. Tebra also sells an AI Voice Assistant for patient calls, quote only.
Example. A private practice that loses about 15 calls a week after hours gets those messages in one queue every morning. A receptionist clears the queue before 9:00. The expected gain is fewer lost new-patient enquiries, which you can count from your own call log.
Watch-out. A call about a medical complaint is patient information, so the vendor needs to sign the agreement your privacy rules require before you route patient calls through it. The assistant must hand any urgent-sounding call to a human or an emergency message, and must never answer clinical questions. The risks and the shortlist are covered in best AI receptionist for a medical practice: what each type does.
6. A pre-submission claim checker
Rejected claims cost rework. A claim checker reviews a claim before submission for missing information, mismatched codes or eligibility problems. Tebra's AI Billing Assistant is one example; its page describes pre-submission claim review and lists no price.
Example. Illustratively, a solo physician's practice submits 400 claims a month and 6 per cent come back for fixable errors: 24 claims at about 20 minutes of rework each, which is 8 hours. A checker that catches half saves about 4 hours a month. Compare four hours of your biller's time with the quote before committing; at this volume a checker built into the billing system you already use is usually the only one worth pricing.
Watch-out. The physician owns the code. A tool that suggests a higher code than the note supports is an upcoding risk, and an audit trail matters. See AI medical coding software: audit trails and upcoding risk before you turn on suggestions.
7. A business chat plan for everything without patient data
The last tool on the list is the least medical: a business plan of ChatGPT, Claude, Copilot or Gemini for the practice's own writing. ChatGPT Business lists at $20 per user a month billed annually or $25 monthly, from two users; Claude's Team plan lists the same $20 and $25 per seat, for 2 to 150 people; Microsoft 365 Copilot Business is $21 per user a month on an annual plan (with an $18 first-year offer shown until 31 December 2026); and Gemini in Gmail, Docs and Meet comes with Google Workspace Business Standard at $14 per user a month.
Example. The practice manager uses it to rewrite the cancellation policy in plain language, to draft a job advert, to turn meeting notes into an action list, and to write the monthly newsletter. No patient information goes in. For a physician reading this alone, a personal plan at about $20 a month does the same work, but the privacy settings differ and it is not the place for patient details.
Watch-out. Business plans do not train on your content by default, but that is not the same as having an agreement for patient information. Keep a rule on the wall: patient details do not go into this tool.
Three jobs worth testing on any of these tools
Whichever tool you pick, test it on tasks that expose its weaknesses, not on an easy visit. These three are illustrative and use no real patients.
The interrupted consultation. Run a visit where the patient changes topic twice and a nurse walks in. Scribes tend to merge threads. Check whether the note keeps the two problems separate and whether the nurse's comment is wrongly attributed to the patient.
The negation test. Say aloud "no chest pain, no shortness of breath, no fever". A good draft records them as negatives in the right section. A weak one drops the "no" or turns the list into a symptom list. This is the single most dangerous slip in a draft, so test it on purpose.
The numbers test. Say a blood pressure of 142 over 88, a heart rate of 76 and a medication dose out loud. Compare each number in the draft against what you said. Dosing errors and transposed numbers are rare but costly, which is why you read the draft against your memory and the chart, not against the audio.
Referral letters and patient sheets: a before and after
Premier-level scribes and clinical assistants can draft letters. Take a plain request, entered with no names: "Write a referral letter to cardiology for a 58-year-old with exertional chest tightness for 6 weeks, normal resting ECG, on a statin and a blood-pressure tablet, asking for an exercise test and review." The raw draft might open "Dear Colleague, thank you for seeing this gentleman...". Two things to fix: the draft assumed the patient's sex because "gentleman" was not in your facts, and it may have invented a family history sentence to make the letter feel complete. Delete both, add the real details from the chart, and sign.
The second use is a patient information sheet. Ask a clinical assistant for a one-page plain-language explanation of a common condition at a reading age of 12, with a "when to call us" box that you then rewrite to match your own safety-net advice. This takes five minutes instead of thirty, and you can reuse it for years once reviewed. Record the review date on the sheet, and look at it again when guidance changes.
Which physicians gain most
The gain differs by practice style. A primary-care physician with 15 to 25 short visits a day gains most from a scribe, since the notes are repetitive and the cost per note is low. A procedure-heavy specialist, who writes long operative or consult notes, may prefer dictation and a template, with a scribe for clinic days only. A physician who runs mostly telephone or video follow-ups can use a scribe on those calls as well, which most tools support. A locum or part-time physician across several sites needs a tool that works on any browser or phone and does not depend on one EHR, which favours a standalone scribe over an EHR-built one. Match the tool to your week, not to a feature list.
A quick sum for one physician
Say a physician writes 18 notes a day, four days a week, and spends about 6 minutes on each: roughly 7 hours a week. If a scribe brings that to 3 minutes of review per note, the saving is about 3.5 hours a week, or around 175 hours a year over 50 working weeks. At Freed Core's $79 a month the annual cost is $948, so each saved hour costs about $5.40. If the saving is only one hour a week, the same plan costs about $19 an hour saved, which may still be fair but is a different decision. These figures are invented for the arithmetic; use your own minutes.
Five ways physicians overpay for AI tools
- Buying the annual plan first. A pilot on a monthly or free plan costs a little more per month but saves a year's commitment to a tool you drop in week three.
- Signing before reading the retention terms. Ask how long audio and transcripts are kept, who can access them and what happens at cancellation. Put the answers in one document.
- Leaving staff on personal accounts. If the front desk uses free chat tools for patient letters, you have a data problem. Give them an approved tool and a rule.
- Treating a vendor's time-saved figure as yours. One vendor reports a self-reported saving of about nine minutes per note. Your saving depends on your visits, your template and your editing habits. Measure it.
- Ignoring the exit. Know how to export your templates and notes before you need to.
What to buy first, and what to leave
For a typical physician the order is the scribe first, the free clinical assistant second, and the chat plan third. Phones and claim checks come later and only if your own numbers support them. A rough picture for one physician, with list prices:
| Package | What it includes | Approx. monthly cost |
|---|---|---|
| Starter | Heidi Free or Freed trial, ChatGPT for Clinicians if eligible | $0 |
| Working set | Freed Core ($79) or Heidi Clinician ($110, annual), plus ChatGPT for Clinicians | $79 to $110 |
| Full | Freed Premier ($104, annual) or Heidi Clinician ($110, annual), plus a phone assistant from about $150 and a $20 to $25 chat plan | About $275 to $285, plus quotes |
The quotes are the unknown: Tebra's tools, a phone vendor's volume pricing and a claim checker will each need a conversation. The numbers above are only useful once you add your own: how many notes you write, how many calls you lose, how many claims bounce.
A one-month test that gives you a straight answer
- Week 0. Log how long notes take, for five days. Count lost calls and rejected claims for the month.
- Week 1. Trial one scribe on a light day. Read every note against what you remember. Count the edits.
- Week 2. Trial the second scribe the same way, or the clinical assistant on three real questions you can verify.
- Week 3. Ask each vendor for the agreement, the retention terms and the exit terms in writing.
- Week 4. Compare minutes saved a day, edits per note, patient reactions and cost. Keep one tool, drop the rest, and diarise a review in three months.
Two habits protect you whichever tool you pick. Read the Plan section of every AI-drafted note line by line, because that is where an invented order or follow-up appears. And tell patients plainly that a tool is listening and that they can say no. A stack for three doctors and four staff shows how the tools fit when more people are involved.
If you would like someone to walk through your own numbers and shortlist with you, that is what my AI implementation consultation covers; get in touch.
Questions physicians ask before paying for AI
Is a free AI scribe good enough for private practice?
It can be, for a pilot or a low-volume week. Before relying on one, confirm the agreement your privacy rules require, how long audio and transcripts are kept, and what the usage limits are. Paid plans mostly add templates, integrations and support, so decide which of those you would use.
Does ChatGPT for Clinicians replace a scribe?
No. It is a clinical assistant for drafting, research and documents, free for verified clinicians where available, with an individual agreement. A scribe listens to the consultation and drafts the note. Many physicians use one of each, but check eligibility and terms on OpenAI's own pages.
Which AI tool saves the most time for one physician?
For most, the scribe, since notes take an hour or more a day. Measure it: log note-completion time for a week before and a week during a pilot. If the saving is under 20 minutes a day, a cheaper tier or a different tool may be the better choice.
What should I never let AI do unsupervised?
Sign notes, order tests, choose treatments, set billing codes, or answer patients about symptoms. The tools draft and suggest; the physician reads, decides and signs. Treat every output as a junior colleague's first draft.
Further reads
- Is an AI Scribe Worth It for a Small Private Clinic? — Test whether a scribe pays for itself at your volume.
- AI Usage Policy for a Medical Practice: Template and HIPAA Rules — A staff policy to adopt before the first pilot.
- AI Scribe Security: Audio, Retention and BAA Vendor Questions — Audio, retention and agreement questions for any vendor.
- How to Roll Out an AI Scribe Without Losing Patient Trust — Introduce a scribe without losing patient trust.
- HIPAA-Compliant AI Assistants for a Medical Office: BAA by Plan — Which assistant plans offer an agreement.
- Medical AI Chatbot for Providers: Clinical Search or Patient FAQ? — Clinical search or patient FAQ: which chatbot for which job.
- AI in Healthcare for Physicians: Personal Tool or Practice-Wide Tool? — A practice-owner decision guide for individual clinician AI use, staff workflows and the agreement boundary.
- How Can AI Help Physicians Finish Charting? A Four-Week Trial — A four-week AI scribe trial for a physician group: baseline minutes per note, two clinicians, a consent script, an error log and a go or no-go rule.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Heidi pricing page, Freed pricing and Front Desk pages, Tebra AI Smart Staff page, Smith.ai AI receptionist pricing page, OpenAI help centre article on ChatGPT for Clinicians, ChatGPT Business, Claude, Microsoft 365 Copilot Business and Google Workspace pricing pages (all checked 4 November 2026).