denture·id
Buyer guide · Planning

How to use AI agents in a dental practice: a 2026 playbook

Where dental AI pays off first, how to work out the return, the HIPAA questions for every vendor, rules for staff chatbot use and clinical AI, and a 30-day rollout plan.

By the Denture-ID Research Desk · updated Sep 30, 2026 · figures pulled live from our database at build time

Reviewed by Francis Brero, VP of AI Strategy & Data Science Expert · Editorial Board

Key takeaways
  • Start with the phone and insurance, not diagnosis. Missed calls and payer calls are high-volume, easy to measure and low clinical risk.
  • The market is crowded and young. We track 78 dental AI tools in six categories. Most do not publish prices, and many launched in the last two years.
  • A BAA comes before the pilot. Only 21 of 78 tools state publicly that they sign a business associate agreement, and none we could find names the language-model provider behind them.
  • For radiographs, check the clearance, not the claim. 8 tools in our database have an FDA clearance on file, and each clearance covers specific findings and image types.

AI in dentistry has moved from demos to daily use. Phones are answered by voice agents, insurance benefits are gathered by bots, notes are drafted while the dentist talks, and radiographs are read a second time by software. The market is young and crowded, and the tools differ most in the things demos skip: integration, privacy and evidence. This playbook covers where to start, what to ask, and how to roll out AI without creating a compliance problem, with data on 78 dental AI tools.

Agents, assistants and tools

TypeWhat it doesExampleControl you need
ToolSuggests; a person decides.Highlighting a suspected lesion on a bitewing.Clinician review of every output.
AssistantDrafts work for a person to approve.A clinical note drafted from the conversation.Sign-off before the note is final.
AgentActs on its own within limits.Answering a call and booking into the schedule; calling a payer and saving benefits.Clear limits, logging, spot checks and a human handoff.

What exists today

CategoryJobPricingHighest on our Data Index
AI receptionists Answer calls, book, reschedule, capture insurance. 7 of 26 publish a price; entry plans from $49, median $299 a month Viva AI, Smith.ai, Dentina.ai
AI scribes Draft notes, perio charts and codes from conversation. 8 of 19 publish a price; entry plans from $39, median $104 a month Scribeberry, Denti.AI Voice, Gain AI
AI insurance agents Verify eligibility and benefits, call payers, prepare claims. 3 of 13 publish a price; entry plans from $150, median $249 a month DIVA, Zuub, SuperDial
AI radiograph analysis Detect and measure findings on X-rays. 4 of 14 publish a price; entry plans from $49, median $94 a month Pearl Second Opinion, Overjet, Denti.AI
AI crown design Propose restoration designs from scans. prices on request SprintRay Cloud Design, Dentbird, 3Shape Automate
AI CBCT segmentation Segment teeth, nerves and bone in CBCT. prices on request WebCeph, Relu Creator

Prices are entry plans of tools in our index; per-provider, per-location and per-minute units are mixed, so compare within a category, not across.

Where to start: high volume, low risk

  1. Missed and after-hours calls. An AI receptionist that books into your system recovers new patients who would otherwise call the next practice. Easy to measure against your call logs. See AI receptionists compared.
  2. Insurance verification. AI insurance agents take payer calls and portal lookups off the front desk. Measure hours saved and verification errors. See insurance verification software.
  3. Clinical notes. AI scribes save time at the end of each appointment but record patient conversations, so they raise the strictest privacy questions. Pilot with one willing provider.
  4. Radiograph AI. Useful as a second reader and for showing patients what the dentist sees. Introduce it once the team is comfortable with AI in the workflow, and use only tools cleared for the findings you rely on.

Check first what your practice management system already includes. Several now build in AI notes, voice perio charting or X-ray analysis, which may remove the need for a separate vendor.

Will it pay for itself?

Work it out before the pilot, with your own numbers. Two illustrations:

AssumptionAI receptionistAI insurance agent
What it recovers each month10 extra new-patient bookings40 staff hours
Value of each unit$150 first-visit collections$25 per hour, loaded
Monthly value$1,500$1,000
Example subscription$349$599
Net per month$1,151$401

All inputs are assumptions for illustration. A new patient is usually worth far more than the first visit over time; staff hours only save money if the time is redeployed or overtime falls.

The inputs that matter are in your data: missed calls per month from your phone system, and hours spent on verification from a week of time tracking. If you cannot measure the baseline, you cannot tell whether the tool worked.

HIPAA checklist for any AI vendor

Across the 78 AI tools we track: 69% claim HIPAA compliance, 27% state that a BAA is available, 4% confirm that a person reviews output, and none that we could find names its language-model provider. A compliance badge on a website is not a BAA. Ask every vendor:

  1. Will you sign a BAA before the pilot starts?
  2. Which subprocessors, including the language-model provider, receive patient data, and do they have BAAs with you?
  3. Are audio, transcripts or images retained? For how long? Are they used to train models?
  4. Where is data stored, and can you provide a SOC 2 Type II report or equivalent?
  5. What access does the integration have in our practice management system, and is it logged?
  6. How does a person review, correct or override the AI’s output?
  7. What happens to our data if we cancel?

Staff use of ChatGPT and other chatbots

The most common AI risk in a practice is not a vendor. It is a team member pasting a patient’s details into a free chatbot to draft a letter or summarize a record. Consumer accounts are not covered by a BAA.

  • Allowed without restriction: tasks with no patient information, such as job adverts, general newsletters, policy drafts and training material.
  • Allowed only on an approved account: anything with patient information, and only on a business or enterprise plan that signs a BAA and turns off training on your data.
  • Not allowed: diagnoses or treatment decisions based on a general chatbot’s answer.

Rules for clinical AI

8 AI tools in our database have an FDA clearance on file. See the FDA clearance register.

  • A clearance is specific. It covers named findings, image types and patient ages. Software cleared for caries on bitewings is not cleared for bone loss on panoramic images. Read the indications, not the press release.
  • The dentist remains responsible for every diagnosis. Record in the chart that AI was used as an aid.
  • Watch for false positives. A second reader that flags too much can lead to over-treatment. Review a sample of flagged and unflagged images in the first month.
  • Vendor accuracy figures are the vendor’s. Ask for the clinical study submitted for clearance and whether any independent study exists.

Our reviews of the main tools, with their clearances checked against the FDA letters: Overjet, Pearl, VideaAI.

Telling patients

There is no general federal requirement to tell patients that AI is used, but states have begun to legislate. California’s AB 3030, in force since January 1, 2025, requires health facilities, clinics and physician offices that use generative AI for patient communications about clinical information to say so and explain how to reach a person, unless a licensed provider reviews the message first. Ask your adviser whether and how it applies to your practice, and check your own state.

Transparency is sensible anyway. A short greeting from a voice agent (“you’re speaking with our virtual assistant; say ‘front desk’ at any time”) and a sentence in your privacy notice avoid surprises.

A 30-day rollout plan

  1. Week 1: choose and measure. One workflow, one owner, a written baseline: missed calls, verification hours or note completion time.
  2. Week 2: contract and connect. Sign the BAA, limit the integration’s permissions, and run in shadow mode where possible: the AI works, a person checks everything before it counts.
  3. Week 3: go live, narrowly. After hours only, one provider or one payer group. Review every AI action daily.
  4. Week 4: decide. Compare with the baseline. Expand, adjust or stop, and write down why.

A one-page AI policy

Put these points on one page and have every team member sign it:

  1. The list of approved AI tools, and who approved each one.
  2. No patient information in any tool that is not on the list.
  3. Every clinical AI output is reviewed by a clinician before it affects care.
  4. Every AI-drafted patient message is reviewed before sending, unless the tool and use have been approved for automatic sending.
  5. Errors and near misses are reported to the tool’s owner.
  6. The list and the policy are reviewed every six months.

Compare tools in the dental AI hub and the AI trust matrix, which shows BAA availability, human review and model disclosure for every tool.

FAQ

How can AI agents be used in a dental practice?

The established uses are: answering and booking calls (AI receptionists), verifying insurance eligibility and chasing claims (AI insurance agents), drafting clinical notes and perio charts from conversation (AI scribes), flagging findings on radiographs (AI radiograph analysis), and proposing crown designs from scans. We track 78 dedicated tools across these categories.

What is the difference between an AI agent and an AI tool?

An AI tool helps a person do a task, such as suggesting a note or highlighting a lesion for the dentist to review. An AI agent carries out a task on its own within set limits, such as calling a payer, reading the benefits and writing them into the patient record, or answering a call and booking an appointment. Agents save more time and need tighter controls.

Do dental AI tools need a BAA?

Yes. Any tool that receives, stores or processes protected health information on the practice’s behalf is a business associate under HIPAA and needs a business associate agreement. Only 21 of 78 AI tools we track state publicly that one is available; ask every vendor before sharing patient data.

Can dental staff use ChatGPT?

Not with patient information, unless the practice has a business account with a signed BAA and data-retention controls. Consumer chatbot accounts are not covered. Staff can use them for tasks with no patient data, such as drafting a job advert or a general newsletter.

Is dental AI FDA approved?

Software that detects or measures disease on radiographs is a medical device and needs FDA clearance, usually through a 510(k). 8 tools in our database have a clearance on file; we have checked those of Overjet, Pearl Second Opinion, and VideaAI against the original FDA letters. Receptionists, scribes and insurance agents are administrative tools and are not cleared by the FDA.

What is the best AI for dentists?

There is no single one. Choose by the problem: an AI receptionist if calls go unanswered, an insurance agent if verification consumes staff hours, a scribe if notes run late, and FDA-cleared radiograph AI if you want a second reader and a patient-education aid. Our rankings for each category are in the dental AI hub.

Do I have to tell patients that AI is used?

Federal rules do not require a general notice, but some states have started to. California’s AB 3030, in force since January 2025, requires health facilities, clinics and physician offices to disclose generative AI in certain clinical communications to patients unless a licensed provider reviews them first. Check your state’s rules, and consider telling patients anyway, for example when an AI answers the phone.

Method & sources

Counts, prices, BAA availability, language-model disclosure and FDA clearances come from our database of dental AI tools; each value links to the vendor or FDA source it was taken from. FDA clearances for Overjet, Pearl and VideaHealth were checked against the original clearance letters. Payback examples are illustrations with stated assumptions. This page is general information, not legal or compliance advice.

Product facts come from the Denture-ID database, where every field links to the vendor page or document it was taken from — open any product profile to see them. Rankings use the Data Index; vendors cannot pay to change a position. Found an error? Tell us.

  1. FDA: artificial intelligence-enabled medical devices — list of cleared AI devices, including dental radiograph software
  2. HHS: guidance on HIPAA and cloud computing — a vendor that processes patient data is a business associate
  3. Future of Privacy Forum: analysis of California AB 3030 — disclosure of generative AI in patient communications
  4. Planet DDS: AI Voice Perio in Denticon — example of AI built into a practice management system

More buyer guides