Dental AI Weekly
Dental AI Weekly: When Dental AI Starts Doing the Work
New healthcare guidance asks what AI should be allowed to do on its own.
Welcome to this week’s Dental AI Weekly, honest analysis of where dental AI is going, from someone building in it.
Most dental AI still waits for you to ask it for something. It flags an image, drafts a note, or answers a question. An AI agent goes a step further. You give it a job, and it can complete several steps on its own.
Three developments this week show the pressure building from different directions: patients are drawing boundaries around AI, model processing is getting cheaper, and healthcare groups are defining where AI agents must stop.
The technology is getting cheaper to run before most practices have decided what it should be allowed to do.
WHAT HAPPENED THIS WEEK
01- WHERE PATIENTS DRAW THE LINE
Patients are more comfortable when AI helps with costs than when it weighs in on a diagnosis.
Humana shared survey findings on July 31. Forty percent of respondents said they were comfortable with AI estimating billing costs. Thirty-six percent were comfortable with AI confirming a diagnosis, while 30% were uncomfortable with it. Humana commissioned the research, which Opinium conducted in July 2025. Sources: Humana News and the Humana infographic
These numbers show what people said they were comfortable with. They do not tell us how many patients use these tools, whether the tools are accurate, or whether they change treatment decisions. Humana did not publish the full questionnaire or sample size.
What this means for your practice: Patients are not equally comfortable with every use of AI. A cost estimate feels different from a machine appearing to confirm a diagnosis. Tell patients what the tool is doing, and make it clear that the dentist is still making the clinical call.
02- AI GETS CHEAPER
OpenAI cut the price of GPT-5.6 Luna by 80%.
On July 30, OpenAI cut the price of GPT-5.6 Luna by 80%. Luna now costs 20 cents per million input tokens and $1.20 per million output tokens. Tokens are the units used to charge for the information an AI model reads and writes. OpenAI also cut the price of its mid-tier Terra model by 20%. Sources: OpenAI and CNBC
The price of the model is only one part of an AI product’s cost. It does not include software connections, voice processing, secure storage, monitoring, compliance work, or the people who review mistakes. Lower model prices do not establish clinical reliability or product quality.
What this means for your practice: This does not make dental AI products 80% cheaper. It changes what software companies can afford to have AI do in the background all day. Recall lists, insurance correspondence, referral intake, eligibility checks, chart review, and multilingual follow-up all become less expensive to process. As the model gets cheaper, the harder parts become more obvious: connecting to dental systems, protecting patient information, catching mistakes, and proving the work was actually completed. Cheap AI that needs constant cleanup is still expensive.
03- AI THAT ACTS
Healthcare is deciding how to test AI that can take action on its own.
On July 29, the Coalition for Health AI released a guide for AI agents in healthcare. An AI agent is software that can take a goal, decide what steps to take, and carry out those steps with less day-to-day direction from a person.
CHAI’s guidance gets specific. As patient information moves between AI tools, the reason it is being used and the patient’s permission should move with it. A person should sign off before AI-added information goes into the health record. If a phone agent hears a clinical warning sign, it should always stop and escalate to a person.
CHAI also says some jobs should not use an agent at all. If a simple, predictable automation can do the job, giving AI more freedom may only add cost and risk. This is guidance, not law or proof that these systems improve dental care.
What this means for your practice: Most dental AI helps with one step today. Future systems may handle a whole chain of work, such as checking coverage, finding missing information, scheduling the patient, and preparing the claim. Before that happens, the practice needs to decide what the software can do without asking, when it must stop, and who checks its work. A wrong answer in a chat is easy to spot. A wrong action moving quietly through the practice can travel much farther.
“A wrong answer in a chat is easy to spot. A wrong action moving quietly through the practice can travel much farther.”
BY THE NUMBERS
Roughly 3x
One health system found roughly three times as many AI agents had been built with approved tools as had gone through the health system’s official review, according to CHAI’s July 29 release.
What it signals: It was easier to build the agents than to review them. Dental practices could face the same problem. Buying from an approved vendor does not tell you what its AI can see, change, send, or add to a patient’s record.
Limitation: This came from one unnamed health system, not a national study. CHAI did not publish the number of agents involved, the dates, or the records behind the finding.
READER Q&A
“I keep hearing that AI scribes improve case acceptance. How does having a microphone in the room actually make a patient more likely to accept treatment?”
Dr. K.S.
BH: A microphone by itself does nothing for case acceptance. Patients do not accept treatment because software was listening.
The benefit comes from what an ambient scribe gives back to the people in the room. When the dentist is not typing, clicking through templates, or mentally reconstructing the clinical note, they can pay closer attention to the patient. They can make eye contact, notice hesitation, answer the actual concern, and explain treatment more clearly.
The same applies to the rest of the team. Better documentation and clearer handoffs reduce the chance that a patient reaches the front desk confused about what was recommended or what happens next.
That can support case acceptance, but it is not automatic. I would not buy a scribe because someone promises it will raise your acceptance rate. Buy it because it removes documentation friction and returns attention to your patients. If your team uses that attention to communicate better and build more trust, improved case acceptance may follow.
With OraCore Scribe, that starts with capturing the visit and drafting the clinical notes so the team can stay present instead of trying to remember everything later. The technology creates the opportunity. The people still create the outcome.
FROM THE ORACORE BLOG THIS WEEK
Practice Economics | 10 min
Where Does AI Actually Show Up Financially in a Small Dental Practice?
Define the job first. This article shows how to measure the workflow an AI tool changes before assigning it credit for revenue, capacity, or growth.
Practice Growth | 3 min
AI for Dental Practice Growth Without Leaks
More activity does not prove that AI improved the visit. Follow the attention, handoffs, and completed treatment to see whether the workflow actually changed.
A good first AI workflow has one clear job, a person checking the result, and time you can see it saving. OraCore Scribe starts with a focused task: capture the visit and draft the clinical note for review.
That’s the week. Reply with what you’re seeing in your own practice. I read every one.
Brad Hutchison
CEO, OraCore AI
Dental AI Weekly drops every Monday. pulse@oracorenews.com | OraCore AI, Denver CO