The short version
Some tool links on this page are affiliate links. If you sign up we may earn a small commission at no extra cost to you. We list tools we would recommend either way. Full disclosure.
- Six tasks where AI helps a medical practice in 2026: clinical documentation, prior authorization, scheduling and no-show reduction, billing and coding, after-hours triage routing, and patient follow-up.
- Start with a low-risk administrative task, not clinical decision support. Scheduling, no-show reminders, and intake carry little downside if the AI errs. Anything that influences a diagnosis or treatment raises patient-safety and FDA-clearance questions and belongs much later, if at all.
- The solo setup: an ambient scribe such as Heidi Health (from $50) or Suki ($99) plus a patient-communication tool runs roughly $150 to $400 a month per provider. A 2-to-5-provider practice adding prior authorization lands in the $400 to $1,500 range.
- Sign a Business Associate Agreement (BAA) before any protected health information touches a tool. Purpose-built medical tools (Nuance DAX Copilot, Abridge, Suki) sign a BAA and do not train on patient data. General-purpose consumer AI does not, and PHI should never go into it.
- Verify before you adopt. Signed BAA, HIPAA Security Rule safeguards, FDA clearance status for any clinical tool, and breach-notification terms are your compliance review to run with your compliance officer, not the vendor's claim to accept. See the checklist below.
Find a local AI pro
What do solo and small-practice physicians ask about adopting AI?
The questions physicians actually put to AI about bringing it into a practice, answered directly.
What does an AI medical scribe do, and is it HIPAA-compliant?
An ambient AI scribe (Nuance DAX Copilot, Abridge, Suki, Heidi Health) listens to the visit with patient consent and drafts a clinical note the physician reviews and signs. Purpose-built medical scribes sign a Business Associate Agreement (BAA), encrypt data in transit and at rest, and state they do not train models on patient data. Confirm the signed BAA and the data-handling policy with your compliance officer before any tool processes protected health information (PHI). General-purpose consumer AI such as ChatGPT does not sign a BAA and should never receive PHI. This is general information, not legal or compliance advice.
Can AI automate prior authorization?
Yes, in part. AI tools draft prior authorization requests by pulling the relevant clinical justification from the chart, submit them to payers, and track status and denials. Staff and the physician review each request before submission. Vendor-reported case studies describe turnaround dropping from several days to under 48 hours (vendor-reported; verify before relying on it). The clinical justification and the appeal decision remain the practice's responsibility, not the tool's.
How does AI reduce no-shows and help scheduling?
AI-driven scheduling tools send automated appointment reminders by text and email, offer patient self-scheduling, and fill canceled slots from a waitlist. Vendor-reported figures put no-show reduction in the 20 to 40 percent range (vendor-reported; verify before relying on it). This is one of the lowest-risk places to start, because an error costs a missed reminder, not a clinical mistake. Any patient-facing message must avoid protected health information unless the channel is covered by a BAA.
Can AI help with medical billing and coding?
Yes, as a suggestion engine the practice reviews. Practice-management tools with AI (Tebra, SimplePractice, and EHR-integrated coding assistants) suggest E/M codes from the documentation, flag likely undercoding, and surface audit risk before submission. The coder or physician confirms every code before it bills. AI narrows the work and reduces missed charges; it does not replace the practice's responsibility for accurate coding and compliant billing.
Are AI patient-triage chatbots safe to use?
Use them only for administrative triage, not clinical advice, unless the tool is FDA-cleared for that purpose. AI chatbots can route patient messages, collect intake information, and direct urgent symptoms to a human or to call emergency services. They should not provide a diagnosis or a treatment recommendation. Any tool that influences a clinical decision raises FDA software-as-a-medical-device questions and liability for AI-assisted decisions. Confirm scope, FDA status, and your malpractice carrier's position before deployment. This is general information, not legal or medical advice.
Which medical AI tools are FDA-cleared, and which are not?
Most documentation, scheduling, and billing tools are administrative software and are not FDA-regulated, because they do not make a clinical decision. Diagnostic and clinical-decision-support tools that interpret images, ECGs, or other clinical data may require FDA clearance, often a 510(k). Check the FDA's database of authorized AI-enabled medical devices for any tool that touches diagnosis or treatment, and confirm the specific cleared indication, not a general marketing claim. A scribe that drafts a note is not the same as a tool that reads a scan.
Which AI tasks are safe in a practice, and which are risky?
Lower-risk tasks: scheduling and no-show reminders, intake collection, ambient documentation a physician reviews and signs, and coding suggestions a coder confirms. Higher-risk tasks: any tool that diagnoses, recommends treatment, or generates clinical content that reaches the patient or the record without physician review. The dividing line is not the task but the human-verification step and the FDA status. Two safeguards cover most of the risk: use tools that sign a BAA for anything touching PHI, and keep a clinician responsible for every clinical output. See Before you adopt any AI tool below.
What does AI actually do for a medical practice?
Four areas across the patient visit: (1) getting found and capturing the inquiry, (2) the visit and documentation, (3) authorization and billing, (4) follow-up and retention. Most practices start with one, confirm the result over a defined pilot, then add a second.
-
01
Get found and capture the inquiry, including after hours
How prospective patients look for a physician has split into two paths, and a practice needs to be present on both:
- Search and maps (still the largest): Patients search Google and read your Google Business Profile, reviews, and website. Visibility there drives most new-patient inquiries.
- AI assistants (newer, growing): Patients ask ChatGPT, Perplexity, or Google's AI for a doctor in their area and specialty. Visibility there depends on how those engines read your site and where your practice is referenced across the web.
AI scheduling and intake tools answer the inquiry on either path, collect the intake information, and book the appointment around the clock instead of routing an after-hours call to voicemail. Keep any protected health information out of channels not covered by a BAA.
Tools: SimplePractice, Tebra, Phreesia.
-
02
The visit and documentation — where AI saves the most physician time
Ambient scribes capture the visit and draft the clinical note. The physician reviews and signs. The evening charting backlog is the single largest time drain AI addresses.
- Draft the SOAP note from the visit conversation, with patient consent
- Pull structured data into the EHR for the physician to confirm
- Cut documentation time per visit, so notes finish during the day rather than after hours
Tools: Nuance DAX Copilot, Abridge, Suki, Heidi Health, DeepScribe.
Vendor-reported figures put note-writing time reduction in the 50 to 70 percent range (vendor-reported; verify before relying on it). The physician remains responsible for every signed note.
-
03
Authorization and billing — drafting and coding the practice reviews
AI handles the drafting and the first-pass coding. Staff and the physician confirm before anything is submitted.
- Draft prior authorization requests from the chart and track payer status
- Suggest E/M codes from the documentation and flag likely undercoding
- Surface audit risk before a claim is submitted
Tools: Tebra, SimplePractice, EHR-integrated prior-authorization assistants.
-
04
Follow-up and retention
Patient follow-up is the work a practice always means to do and rarely has time for. AI does it on a schedule.
- Automated appointment reminders reduce no-shows
- Post-visit follow-ups and care-plan check-ins keep patients engaged
- Recall reminders bring patients back for annual visits and screenings
Keep protected health information out of any reminder or follow-up channel that is not covered by a BAA.
Which AI tools work for medical practices?
Pricing reflects published vendor information as of May 2026. Verify current pricing, BAA availability, and FDA status directly with each vendor before purchase.
| Tool | Category | Best for | Starting price | Key constraint | Setup time |
|---|---|---|---|---|---|
| Heidi Health | Ambient scribe | Solo and small practices testing a scribe | From $50/mo | Confirm signed BAA before PHI | Same day |
| PatientNotes | Ambient scribe | Specialty clinics needing non-primary-care templates | From $50/mo | Confirm signed BAA before PHI | Same day |
| Suki | Voice assistant + scribe | Practices on Epic, Cerner, or athenahealth | From $99/mo | EHR integration required; confirm BAA | 1-2 weeks |
| DeepScribe | Ambient scribe | Multi-specialty practices wanting deep EHR integration | ~$99-149/mo | Confirm signed BAA before PHI | 1-2 weeks |
| SimplePractice | Practice management + intake | Solo and small practices wanting one system | From $49/mo | Confirm BAA on the plan you buy | 1-2 weeks |
| Tebra (formerly Kareo) | Practice management + billing | Independent practices needing billing and scheduling | Custom | Confirm BAA and coding scope | 2-4 weeks |
| Nuance DAX Copilot | Ambient documentation Larger Practices | Mid-size and larger practices on Epic | Custom | Enterprise rollout; deep Epic integration | Multi-week |
| Abridge | Ambient scribe Larger Practices | Multi-site practices and health systems | Custom | Enterprise rollout; confirm BAA | Multi-week |
A solo physician or small practice should start with an ambient scribe (Heidi Health or Suki) for the documentation burden, then add scheduling or prior authorization within 60 days. Nuance DAX Copilot and Abridge are built for mid-size and larger practices with EHR integration teams; they are usually a heavier fit than a small practice needs. Confirm a signed BAA and the FDA status of any clinical feature before adoption.
What does an AI setup actually cost for a medical practice?
Real monthly bundles by practice size, based on published vendor pricing as of May 2026. HIPAA-compliant vendors cost more than general AI; verify each tool's current pricing and BAA terms before purchase.
| Practice size | Tools | Total per month | Setup time |
|---|---|---|---|
| Solo physicianyou + front desk | Heidi Health (from $50) + patient communication ($50-150) | $150-$400/mo | Same day - 2 weeks |
| Small practice2-5 providers | Suki or DeepScribe (per provider) + Tebra billing + patient comms | $400-$1,500/mo | 2-4 weeks |
| Mid-size practice6-15 providers | EHR with AI + Abridge or Nuance DAX (per provider) + prior-auth automation | $2,000-$6,000+/mo | 4-8 weeks |
| Larger group16+ providers, multi-site | Nuance DAX Copilot Enterprise + prior-auth and denials automation | $3,000-$15,000+/mo | Multi-week |
Enterprise scribe pricing is quoted per provider per month and is often packaged with EHR vendor agreements. For most independent physicians, the solo $150-$400 ambient-scribe bundle is the common starting point, with the break-even reached on the charting time recovered in the first month.
A week with AISee what a typical week with AI might look like in a small medical practice → +
Here is what a typical week could look like for a two-physician primary-care practice running an ambient scribe such as Nuance DAX Copilot, a scheduling and patient-communication tool, and a prior-authorization assistant. All vendors have signed BAAs on file. Hypothetical illustration; results depend on practice size, payer mix, and how consistently the team uses the tools.
Monday morning. 18 visits are scheduled. The scheduling tool sent reminders 48 hours and 4 hours ahead. 17 of 18 patients arrive and one reschedules in advance by text, instead of a no-show. The front desk did not make a single reminder call.
Monday afternoon. The ambient scribe listens to each visit, with patient consent, and drafts the note in real time. The physician reviews and signs after each patient, a couple of minutes per note rather than reconstructing the day's charts after hours. Documentation is finished before the physician leaves.
Wednesday. A patient needs prior authorization for an MRI. The prior-authorization assistant drafts the request with the clinical justification pulled from the chart. Staff review the draft for accuracy and submit it. Approval comes back inside two days instead of the usual several.
Thursday. The coding assistant flags two visits that were likely undercoded based on the documentation. The biller reviews each one, confirms the supporting note, and corrects the code before the claim goes out. Nothing bills without a human confirming it.
Friday. Post-visit follow-up messages and recall reminders go out automatically, with no protected health information in the message body. Several patients book their annual visit. The physician spends 45 minutes reviewing the week's dashboard rather than catching up on unfinished notes.
None of this replaces the physician's clinical judgment. AI drafts the note, the authorization, and the reminder. The physician reviews, verifies, and signs everything that touches a patient's care.
DIY or hire a local AI consultant?
Both paths work. The right one depends on time, in-house technical and compliance capacity, and how much disruption the practice can absorb. Click the path that fits.
DIY makes sense if...
- You or someone on staff is comfortable with new software
- Your compliance officer can review BAAs and security documentation
- The practice can absorb the setup time over a defined pilot
- You are adding one AI tool at a time
- You have run at least one prior EHR or software migration
Hire a local AI consultant if...
- You want to add two or more AI tools in the same year
- You have not run a vendor compliance review before
- Time is the constraint, not budget
- You want someone who has done this in other medical practices
- You want help confirming BAA terms and FDA status up front
A typical local AI consultant for a medical practice will quote you on a flat-fee or retainer basis. The consultant does not replace your compliance officer; verify the consultant's experience handling protected health information before engaging.
How do I start using AI in my medical practice?
A solo physician or small practice can run through these steps over a defined pilot. The first step is the compliance threshold, not the technology.
5 stepsSee the DIY plan for medical practices → +
- Sign a Business Associate Agreement before any PHI is involved
A signed BAA is the threshold question. If a vendor will not sign one, do not let the tool process protected health information. Confirm the BAA in writing with your compliance officer before anything else. General-purpose consumer AI does not sign a BAA; never paste patient data into it.
- Start with a low-risk administrative task
Begin with scheduling, no-show reminders, or intake automation. These carry little downside if the AI errs. Do not start with clinical decision support, which carries patient-safety and FDA-clearance considerations. Pick the administrative task taking the most staff or physician time each week.
- Run a defined pilot and measure one metric
Roll the tool out for 30 to 60 days on a subset of visits or patients. Measure the specific thing you wanted to fix: documentation time per visit, no-show rate, or prior-authorization turnaround.
- Keep a physician in the loop on every clinical output
AI drafts; the physician reviews and signs. No AI-generated clinical note, code, or patient-facing communication leaves the practice without a clinician's review and verification.
- Measure, then either expand or change tools
After the pilot, check the metric. If documentation time or no-show rate moved, expand to a second workflow such as prior authorization. If it did not move, change the tool rather than the category.
I was finishing notes at 9 PM most nights. We brought in an ambient scribe and ran it for a month before rolling it out. My notes are now done before the patient leaves the room, and I review and sign each one. The compliance review up front was the part that mattered most.
Industry pattern, paraphrased from coverage of medical practice operations, 2024–2025.
We automated the drafting side of prior authorization across the practice. The same volume now goes through with fewer staff hours, and our average approval time dropped from over a week to under two days. Staff still review every request before it is submitted.
Industry pattern, paraphrased from coverage of medical practice operations, 2024–2025.
Before you adopt any AI tool in your practice
The Agentic Index lists AI tools for discovery only. We do not vet vendors, verify security claims, or confirm regulatory compliance. Before adopting any AI tool, verify the items below directly with the vendor and your compliance officer. The listing of a tool here is not an endorsement, a security assurance, or a compliance clearance.
Your own compliance and clinical review is the control, not the vendor's marketing. At a minimum, that review should cover:
- Signed Business Associate Agreement (BAA) before any PHI touches the tool. A written, signed BAA is the threshold requirement under HIPAA before any tool processes protected health information. "It is in our terms of service" is not a BAA. If a vendor will not sign one, the tool does not touch PHI.
- HIPAA Security Rule safeguards. Confirm encryption in transit and at rest, access controls, audit logging, and the vendor's documented administrative, physical, and technical safeguards. Ask for a SOC 2 Type II or HITRUST report.
- FDA clearance status for any diagnostic or clinical-decision tool. Administrative tools (documentation, scheduling, billing) are generally not FDA-regulated. Any tool that interprets clinical data or influences diagnosis or treatment may require FDA clearance, often a 510(k). Verify the specific cleared indication in the FDA's database, not a marketing claim.
- Breach-notification obligations. Confirm the vendor's breach-notification commitments and timelines and how they align with the HIPAA Breach Notification Rule and your state's requirements. Know who notifies patients and when.
- Data residency and data-training policy. Confirm where patient data is stored, who can access it, retention and deletion terms, and that your inputs are not used to train models. Read the data-handling policy, not the homepage claim.
- EHR integration security. If the tool connects to your EHR (Epic, athenahealth, eClinicalWorks, Cerner/Oracle, NextGen), confirm how the integration authenticates, what data it can read and write, and that the connection is covered by the BAA.
- State medical board AI guidance. Several state medical boards have issued or are developing guidance on AI use in clinical practice. Confirm the current guidance in your state and that your intended use is consistent with it.
- Malpractice and liability for AI-assisted decisions. Confirm with your malpractice carrier how AI-assisted documentation or decision support is treated under your policy. The physician remains responsible for every clinical decision; the tool is a draft, not a clinician.
- Patient consent for AI use. Confirm whether your jurisdiction or your tools require patient consent for ambient recording or AI processing of a visit, document the consent, and follow applicable recording-consent rules.
This is general information about areas your review should cover. It is not legal, compliance, or medical advice and is not a substitute for your own compliance officer, your malpractice carrier's guidance, or current guidance from HHS, the FDA, and your state medical board. Review the current rules that apply to your practice and jurisdiction before deploying any tool. Listed AI consultants are likewise not vetted by The Agentic Index for HIPAA Security Rule compliance, BAA practices, or PHI handling; request a signed BAA from any consultant before they touch protected health information.
How do I find a local AI pro for my medical practice?
Tell us your area, your practice size, and your biggest bottleneck. We will route you to a local AI consultant who works with medical practices.
Listings are for informational purposes only. The Agentic Index does not endorse, certify, or vet any provider for HIPAA Security Rule compliance, BAA practices, or PHI handling. Always verify a consultant's credentials and experience handling protected health information, and request a signed BAA, before engaging.
Sources
- Vendor published pricing pages reviewed 2026-05-29 — heidihealth.com, patientnotes.ai, suki.ai, deepscribe.ai, simplepractice.com, tebra.com, nuance.com, abridge.com
- U.S. FDA — Artificial Intelligence-Enabled Medical Devices list for clearance status of clinical tools
- HHS Office for Civil Rights — HIPAA guidance on BAAs, the Security Rule, and breach notification
- Documentation-time and no-show reduction figures: vendor-reported customer case studies, 2024-2025 (vendor-reported; verify before citing)
Last reviewed: 2026-05-29. The Agentic Index does not provide legal, compliance, or medical advice. Verify all claims, pricing, BAA terms, and FDA status directly with each vendor and your compliance officer.