What AI can and can’t do for a dental practice
Last updated October 17, 2026 · Prices and features change, so check the linked sources before you buy.
On October 16, 2026, Dentaltown founder Dr. Howard Farran posted “The AI Bubble and Dentistry,” comparing today’s AI spending to the railroad boom before the Crash of 1873. His point: the railroads were real and so was the financial excess. His test for practice owners is simple: “Does this specific tool improve this specific workflow enough to justify its cost? … A great demo is not ROI.” [1]
We agree. Here’s that test applied to the places AI is actually showing up in practices.
What AI can do well today
1. Answer the phone when your team can’t
This is the most mature category, and there are many vendors. Several (Dentina, Arini, Enamly) say they book straight into your PMS. Others capture requests for your team to confirm (compared here). The ROI test is easy to run: count unanswered calls by hour for two weeks, then measure how many after-hours calls turn into booked appointments during a 30-day pilot.
2. Act as a second reader on radiographs
FDA-cleared detection tools help dentists find things. In a 2025 randomized controlled trial with 30 dentists, AI assistance raised accuracy on periapical radiolucencies from 91.6% to 93.3% and cut false positives from 746 to 345 [2]. In that setting it worked as a brake on overdiagnosis, not an accelerator.
3. Take on documentation and admin
Voice charting and AI notes are spreading. Overjet, for example, has extended voice charting to hard tissue and perio [3]. The value is in staff minutes saved per day, so measure it that way.
What AI can’t do (and shouldn’t be sold as doing)
- Diagnose or treatment-plan on its own. Farran’s August column on Pearl and Joanna Stern sets out the line: AI detection, clinical diagnosis and treatment planning are “not the same act.” Pearl’s clearance is as computer-aided detection, a second reader, and it was not cleared to diagnose periodontitis or prescribe SRP [2].
- Guarantee the same accuracy in your office. In one 2026 independent test, overall caries accuracy was 91.6%, but positive predictive value was 64.7% for any approximal caries and around 40% when grading lesion depth [2]. FDA clearance is not a guarantee of performance on your sensor and your patients. Ask how the tool has been validated on independent data.
- Replace judgement about incentives. Farran calls AI an amplifier. In a conservative culture it may reduce missed disease. In a production-quota culture the same tool becomes a sales aid [2].
- Answer patients from information it doesn’t have. Whether it’s a phone bot or a website bot, if it isn’t grounded in your actual policies it will improvise. A Canadian tribunal held Air Canada liable for exactly that (the case).
A five-question ROI test for any dental AI tool
- Which single workflow does it change, and who on the team owns it?
- What number moves (calls answered, no-shows, minutes per note, claims reworked), and what is it today?
- What will it cost in total over 12 months: setup, per-unit fees, staff training time?
- What happens when it’s wrong, and who catches it?
- Can we pilot it for 30 days and walk away?
The ADA’s AI standards work is a useful reference for validation questions [4]. Our free AI front-desk checklist turns the front-office part of this into a one-page demo script.
FAQ
Is AI in dentistry just hype?
No, but the hype and the value are different things. Howard Farran compares it to the railroad boom: the technology was real and so was the financial excess. Judge each tool by whether it measurably improves one specific workflow enough to justify its cost.
Can AI diagnose cavities or periodontitis?
No. Radiograph AI tools such as Pearl Second Opinion are cleared as computer-aided detection, a second reader. Diagnosis and treatment planning remain the dentist’s job, combining the finding with the clinical exam, history and risk assessment.
Does dental AI lead to overtreatment?
It can go either way. A 2025 randomized trial found AI assistance reduced false positives on periapical radiolucencies and dentists trended more conservative. Farran warns that AI amplifies whatever incentives exist in a practice, so the workflow and culture around it matter.
Where should a small dental practice start with AI?
Usually at the front desk: answering missed and after-hours calls, because the before-and-after numbers are easy to measure. Pilot for 30 days, starting with after-hours and overflow calls, and review transcripts weekly.
Sources
- Howard Farran on X: “The AI Bubble and Dentistry, What the Crash of 1873 Can Teach Dental Practice Owners” (Oct 16, 2026)
- Dentaltown, Howard Farran: “Dental AI, Joanna Stern, Pearl, and the Risk of Overtreatment” (Aug 14, 2026), with the studies it cites
- Overjet: Voice hard tissue and perio charting
- ADA: Artificial intelligence in dentistry (standards)