AI Tools for Medical Practices: A Stack for Three Doctors and Four Staff

A practice with three doctors and four staff should start with an AI scribe for the doctors, then phone handling for the front desk, then claim checks before billing. At list prices the first two cost from $0 on free tiers to roughly $1,000 a month at full coverage, while billing checks are usually quoted. Check each vendor's page first.

The order matters more than the brands. A scribe is the safest first tool because a doctor reads and signs every note. Phones are second because they touch patients directly and need scripts, call routing and an exit plan. Billing checks come last because they only help when the notes feeding them are already clean.

Follow me on Instagram@sagnikteaches

The practice in this example

This is an illustration, not a real clinic. Three doctors each see about 20 patients a day across four days a week. Four staff share the front desk: two receptionists, one practice manager who also handles billing, and one nurse or medical assistant. They use an electronic health record already, and the doctors each spend around 90 minutes a day finishing notes, often at home.

Connect on LinkedInSagnik Bhattacharya

The practice owner wants three things: notes finished before the doctors leave, fewer missed calls at lunchtime and after hours, and fewer rejected claims. The budget is modest, and nobody has time to run a project. That determines the stack: three tools, each with one owner, each piloted before it is paid for in full.

Subscribe on YouTube@sagnikteaches

Tool one: the scribe, three seats and what they cost

An ambient scribe listens to a consultation, with the patient's consent, and drafts the note. The doctor edits and signs. The tools below are the ones named in current vendor pricing, and all figures are list prices checked on the vendors' pages in November 2026. They can change, so treat them as a starting point.

Do the arithmetic with the doctors' volumes, not the plan names. Three doctors at about 80 patients a week each is 240 notes a week, or roughly 1,000 a month. Freed Starter's 40 notes a month per clinician covers about two days. So the realistic scribe choices are Core or Premier from Freed, Clinician or Practice from Heidi, or the free Heidi tier for a pilot. For the three doctors the annual cost runs from $0 to about $6,480 (Heidi Practice at $540 a month); most paid options sit between $2,844 (Freed Core) and $3,960 (Heidi Clinician). Heidi's Practice plan costs more per seat, but its free assistant users matter if the nurse or a receptionist would otherwise need a seat to tidy letters.

A simple test of value: if a doctor saves 45 minutes a day on notes, across four days that is three hours a week, or about 13 hours a month. At $79 to $180 a seat a month, a paid plan needs to save only a small share of those hours to cover itself. But measure it. Have each doctor log note-completion time for a week before the pilot and a week during. For the decision between ambient tools and plain dictation, AI dictation software for doctors sets out when each fits.

What the pilot must check:

• Agreement. Ask for the vendor's business associate agreement (BAA) or the equivalent your privacy rules require, before any patient audio is recorded. Get it in writing.
• Retention. Freed's pricing page says audio is saved temporarily until summaries and quality checks finish, then deleted. Ask every vendor the same question and record the answer. Vendors change defaults, so recheck each quarter.
• Consent. Patients need to be told, and some will refuse. Decide how a doctor turns the scribe off, and put a sign in the room.
• Edit rate. Have each doctor count how many notes need major rewriting in the first week. If it is more than one in five, change the template before judging the tool.

Picking a scribe by how well it fits your EHR is covered in the AI scribe decision rule.

Tool two: phones for the front desk

The front desk is where one missed call becomes a missed appointment or an unhappy patient. An AI phone assistant can answer when the lines are busy, take messages, answer opening-hours questions and book slots that the schedule allows. For a medical practice, the decisions are narrower than for a plumber: what it may say, when it hands to a person, and how it treats a patient who describes symptoms.

Prices vary widely and many vendors quote by the minute or the call. Two published examples: Smith.ai lists its AI receptionist as free for 25 calls a month, then $3 a call, with a Pro plan from $150 a month for 75 calls and larger tiers at 150 and 300 calls. Freed, the scribe vendor, lists Front Desk, an AI medical receptionist in beta for single-site practices with one or two phone numbers, at $149 a month per block of 250 calls, with warm transfer to staff during business hours. These are yardsticks, not recommendations: confirm that any vendor you use for patient calls will sign the agreement your privacy rules require, because a call recording about a medical complaint is patient information. Tebra, which sells an AI Voice Assistant for patient calls inside its platform, does not list a price on its AI page, so you would ask for a quote.

Let's price it with illustrative call volumes. Say the practice receives 60 calls a day on four days, 240 a week, of which about a quarter (60 a week) ring while lines are busy or after hours. If the AI takes only those, that is about 260 calls a month. On Smith.ai's free plan that would be about $705 a month (235 calls at $3), while its 300-call Pro tier is listed at $1.67 a call, about $500. Freed's Front Desk would need more than one 250-call block, so ask whether the extra calls mean a second $149 block or an overage rate. This is exactly why you ask for a price on your volume, not the entry rate. For planning, assume $0 to $150 for a pilot on overflow calls only and $300 to $500 for full after-hours coverage, then check the vendor's page.

Write the call rules before you choose the tool:

• Anything that sounds urgent (chest pain, breathing trouble, a child unwell, bleeding) goes straight to a human or an emergency message, with no AI triage attempted.
• The assistant never gives medical advice, test results or medication information.
• It can give opening hours, directions, the booking link and the repeat-prescription process.
• Every after-hours message lands in one queue that a named receptionist clears first thing.

A realistic mistake: an AI phone agent answers "do I need to come in for this rash?" with reassurance because a script allowed "general guidance". Keep the script to logistics. Best AI receptionist for a medical practice: what each type does compares the types before you shortlist.

Tool three: patient messages and billing checks

The third tool is often two small jobs. The first is patient messaging: appointment reminders, pre-visit forms, and replies to routine questions. The second is claim checking before submission: catching a missing modifier, a mismatched diagnosis or an eligibility problem before the payer rejects it.

Start with messaging, because it may already be paid for. Before buying a separate tool, check what your EHR and your phone tool include: Freed's Front Desk, for example, lists two-way SMS from the same inbox as the call messages, so a receptionist can confirm a detail by text without a second system. Keep messages to logistics (times, forms, directions) and route anything about symptoms or results to a person.

For claim checks in a seven-person practice, the cheapest sensible route is the one inside the system you already use. If your EHR or billing platform has an AI claim-review add-on, ask for the price per claim or per provider and a trial on last month's rejected claims. Tebra, for example, describes an AI Billing Assistant that reviews claims before submission, with pricing on request. Billing tools inside an EHR keep the data in one place and avoid another vendor agreement.

A before-and-after, as an illustration. Before: the practice manager submits around 900 claims a month, and about 9 per cent come back rejected for fixable errors (81 claims). Each rejected claim costs about 15 minutes of rework. That is roughly 20 hours a month. After: a pre-submission check that catches half of those cuts it to about 40 rejected claims and 10 hours. You would value those saved hours at your own rate, and compare with the quote. Without your own rejection rate, nobody can tell you whether this tool is worth it, which is why it sits third.

For choosing between scrubbing, coding and denial tools, read best AI tool for medical billing: scrubbing, coding and denials. If a tool suggests billing codes, the doctor and the biller must still agree them, and the tool should keep an audit trail of every suggestion it made.

The monthly total, three ways

Here are three illustrative packages for the seven-person practice. Prices are list prices at the time of writing and exclude any tax.

Hidden costs to add: an hour or two of setup per doctor, a training session for the front desk, the practice manager's time reading agreements, and the time a doctor spends editing notes. Put a monthly review in the practice manager's diary, with one question per tool: is it saving time, and what went wrong?

Before and after: one consultation note

The scribe is easier to judge when you see what it produces and what the doctor fixes. Here is an illustrative, invented exchange, shortened. The visit: a patient with a persistent cough for three weeks.

Scribe draft:
Subjective: 45-year-old with productive cough for 3 weeks, no fever. Reports mild wheeze at night.
Objective: Chest clear on auscultation. Oxygen saturation 98 per cent.
Assessment: Post-viral cough, likely.
Plan: Reassured. Review in 2 weeks. Chest X-ray ordered.

The doctor reads it and spots two problems. The patient said the cough had been "dry for the first week, then productive", and the draft flattened that. And the doctor did not order an X-ray: they said they would consider one if the cough persisted. The first is a nuance, the second is a fabrication-by-tidying, and the second is the one that matters. The signed note says "X-ray to be considered at review if no improvement".

The lesson for the practice policy is that the doctor signs every note after reading the Plan section line by line, because the Plan is where invented actions appear. Teach that explicitly in the first week. A useful habit during the pilot: each doctor counts how often the Plan needed correcting. If it is more than occasionally, change the template to say "list only actions the clinician states aloud".

Consent wording and the sign in the room

Patients should be told that an AI tool will listen and write a draft, that the doctor reviews it, and that they can say no. Put the wording in three places: a line on the booking confirmation, a sign at the desk and the doctor's spoken sentence. An illustrative spoken line: "I use a tool that listens and drafts my notes so I can look at you instead of the screen. I check everything it writes. Are you happy for me to use it today?" If the answer is no, the doctor switches it off and types. Record the refusal rate for the first month. A rate that is far higher than you expected usually means the introduction is rushed or the sign is missing. How to roll out an AI scribe without losing patient trust goes through this in more detail.

Smaller jobs for a general assistant, kept away from patient data

Once the three main tools are in place, a general assistant on a business plan helps with work that carries no patient information at all. Four examples with sample prompts:

• Policy wording. "Rewrite our missed-appointment policy in plain language at a reading age of 12, 120 words, friendly but clear." The practice manager edits and approves.
• Rota gaps. Paste anonymised shift patterns (initials only) and ask for a cover plan that keeps one nurse and one receptionist on every session.
• Staff meeting notes. Paste your own bullet notes and ask for an action list with owners and dates.
• Template letters. Ask for a blank referral-letter template with merge fields, which the scribe or EHR then fills.

For all four, the rule is that no patient name, number or condition goes into the prompt. If a task needs patient detail, it goes through a tool and an agreement that allow it, or it stays with the humans.

Who owns each tool

A tool with no owner decays. Give each one a named person and a monthly ten-minute check. The scribe belongs to the lead doctor, who watches the edit rate and template quality. The phone tool belongs to the senior receptionist, who reads the overnight message queue and logs wrong answers. The billing check belongs to the practice manager, who compares the rejection rate month on month. The practice owner holds the agreements and renews or cancels. Write the four names on one page and pin it where the team can see it, so that a vendor email about a changed privacy default reaches someone who will read it.

A rollout over eight weeks

1. Week 1. Adopt a one-page AI policy: what tools are approved, what data may enter them, who owns each tool. A template is in AI usage policy for a medical practice.
2. Weeks 2 to 4. One doctor pilots the scribe. Record note-completion time before and during, and the edit rate. Collect the agreement and retention answers in a single document.
3. Week 5. The other two doctors start. Adjust templates for each specialty or style.
4. Weeks 6 to 7. Pilot the phone tool on after-hours calls only. A receptionist reads every message each morning and logs errors.
5. Week 8. Review the numbers. Decide on billing checks only if the rejection rate and the quote justify it.

A good test of whether to continue is plain: has each doctor finished notes earlier, has the front desk missed fewer calls, and has anyone found a patient-safety problem that the tool hid? If the third answer is yes, stop and fix that first. If the answers are good, write the eight weeks up as the practice's standing plan for the next tool.

ChatGPT for doctors, and why it is not the first tool

OpenAI offers ChatGPT for Clinicians, which its help centre describes as free at launch for clinicians who pass its licence verification (physicians, nurse practitioners, physician assistants and pharmacists), open only in its launch market so far, with each clinician able to review and sign an individual agreement under Settings > Agreements. It suits cited clinical questions and drafting letters for one doctor, but it does not replace a scribe tied to the consultation, and an individual agreement does not cover the rest of the team. For one agreement across several users and central admin controls, OpenAI points organisations to ChatGPT for Healthcare. For a practice that wants shared administration, consider it a doctor-level extra, not the stack's foundation. Check eligibility, the agreement terms and any change in availability on OpenAI's own page before relying on it.

If you would rather decide the order with someone who has seen the trade-offs, that is what my AI implementation consultation covers; get in touch. Otherwise, begin with the scribe pilot, count the minutes, and let the numbers choose the next tool.

Questions about the first AI tools for a small practice

Which of the three tools should a practice start with?

Start with the scribe, because it saves the doctors time every day and the risk is easy to contain: a doctor reads and signs every note. Add phones second, once a pilot of the scribe has run for a month. Billing checks come third because they need clean notes and an agreed coding process.

Do all three doctors need the same scribe?

Not necessarily, but one product is simpler to support and cheaper to administer. Let each doctor trial it for a week, since styles differ. If one prefers a different tool, weigh the extra contract and agreement against the benefit before allowing two.

Is a free scribe plan good enough for a practice?

A free tier can suit a pilot or a low-volume doctor. For a practice, check three things before relying on it: the agreement your privacy rules require, how long recordings and transcripts are kept, and whether usage limits will run out mid-week. Free plans change, so confirm them on the vendor page.

Can the front desk use ChatGPT for patient letters?

Only on a plan and agreement that your privacy rules allow for patient information. Do not assume a personal plan has one, and note that a doctor's individual ChatGPT for Clinicians agreement does not cover staff. Without the right agreement, limit staff use to text with no patient details, such as template wording and policies.

Further reads

• AI Scribe Costs Compared: What Small Clinics Pay in 2026 — Compare what small clinics pay for scribes.
• What to Ask Before Buying Any AI Tool for a Medical Practice — Questions to ask before buying any AI tool for a practice.
• HIPAA-Compliant AI Assistants for a Medical Office: BAA by Plan — Which assistant plans offer an agreement, plan by plan.
• HIPAA-Compliant AI Receptionist: A BAA and Retention Checklist — An agreement and retention checklist for AI phone tools.
• Is an AI Scribe Worth It for a Small Private Clinic? — Test whether a scribe pays for itself at your volume.
• Do Small Medical Practices Need an AI Consultant? — Decide whether you need outside help at all.
• AI Medical Coding Software: Audit Trails and Upcoding Risk — Audit trails and upcoding risk before turning on code suggestions.
• A Worked AI Implementation Plan for a Small Medical Practice — A longer implementation plan for a small practice.
• AI Chatbot for a Doctor's Website: Scope, Disclosure, Handover — A practice owner's set-up guide: scope sheet, disclosure line, hand-over rules, logging and named chatbot tools with costs.
• AI Marketing for Medical Practices Without Making Health Claims — Service pages, review replies, referrer updates and local search for a medical practice, drafted with AI and kept free of health claims.
• AI Tools for Physicians: Seven Worth Paying For in Private Practice — Seven AI tools for a private-practice physician, scribes, clinical assistants, phones and claim checks, with what each does and what to verify.
• AI in Healthcare Billing: Patient Statements and Payment Plan Wording — Draft statement and payment plan wording with merge fields, reconciliation checks and a privacy-safe review route.
• 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.
• 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 (all checked 4 November 2026).