Dental AI Weekly
Dental AI Weekly: The Numbers Say Health AI Is No Longer Theoretical
300M weekly health questions, 75% agent resolution, August AI Act dates, and what dentists should check.
Welcome to this week’s Dental AI Weekly, honest analysis of where dental AI is going, from someone building in it.
Patients are already using AI to understand their health.
Some of that will show up in dental visits. A patient may bring a ChatGPT summary, a question about a medication, or an assumption your team needs to verify.
This week’s signals show how AI is changing patient expectations, workflow automation, and the rules for human control.
WHAT HAPPENED THIS WEEK
01- HEALTH CONTEXT
OpenAI moved ChatGPT closer to the patient record.
OpenAI rolled out Health in ChatGPT on July 23 for U.S. users 18 and older across web and iOS. Users can connect Apple Health, supported medical records, One Medical, Function Health, and wellness apps that share data through Apple Health. OpenAI says connected health data and conversations using it are not used to train foundation models or target ads. Source: OpenAI
The bigger signal is patient behavior. OpenAI said more than 300 million people turn to ChatGPT with health-related questions each week. It also said that among early Health users, more than 70% of health-related conversations happened outside the dedicated Health space, which means people want health context inside normal questions, not parked in a special software lane.
What this means for your practice: The first visible change may not be a dental AI product. It may be patients arriving with better language, worse assumptions, or a machine-generated version of their story. Your team needs a simple way to handle that without making the patient feel foolish or letting the AI become the clinician.
Practice check: Write a one-sentence policy for AI-generated patient summaries. Example: patient-provided AI summaries may inform the conversation, but clinical facts must be verified against the patient, the exam, imaging, medical history, and source records before they enter the chart.
02- AGENT WORKFLOWS
OpenAI showed what controlled AI agents are starting to look like.
OpenAI introduced Presence on July 22 as an enterprise product for putting voice and chat agents into customer and internal workflows. Its examples include billing issues, insurance claims, support requests, policy rules, approval steps, and human escalation. OpenAI says each deployment starts with one specific job and limits the agent to the knowledge and system access required for that job. Source: OpenAI
OpenAI also reported that Presence now resolves 75% of inbound issues in its English-language phone support channel without human assistance, with a 15 percentage-point reduction in human handoffs over 10 days. Treat that as vendor-reported data, not independent proof, but it is still a useful operating signal.
The point is not that dental offices need an enterprise call-center agent tomorrow. The point is that AI products are moving from “answer the question” to “handle a bounded workflow with rules, limits, logs, and handoffs.”
What this means for your practice: Do not evaluate an AI agent by asking whether it sounds smart. Ask what job it is allowed to complete. For a dental practice, that might eventually mean checking a benefit, preparing a follow-up, routing an insurance question, or flagging a task for the right person. The useful agent is not the one that acts everywhere. It is the one that knows its boundary and hands off cleanly when a human needs to decide.
Practice check: When a vendor says “agent,” ask for the workflow boundary in writing. What can the AI do by itself? What requires human approval? What system access does it have? What log proves what happened?
03- GOVERNANCE
New AI rules are putting control and audit trails at the center.
The European Commission says the AI Act becomes fully applicable on August 2, 2026, with exceptions. Prohibited AI practices and AI literacy obligations already applied from February 2, 2025. Governance rules and general-purpose AI model obligations applied from August 2, 2025. Some high-risk rules move later: December 2, 2027 for certain high-risk areas and August 2, 2028 for systems embedded into regulated products. Source: European Commission
In the U.S., Representatives Ted Lieu and Nathaniel Moran introduced the AI Kill Switch Act on July 23. The bill would require covered developers to maintain the technical ability to throttle, suspend, or fully shut down covered AI systems. It would also require incident reporting and preservation of forensic records after failures. Source: Congressman Ted Lieu
What this means for your practice: If AI touches patient communication, records, claims, imaging, scheduling, or clinical recommendations, the practice still owns the outcome. That means the office needs a way to know what the AI did, who reviewed it, when a human stepped in, and what record proves it. Control is not a legal detail for later. It is part of the workflow.
Practice check: Pick one AI workflow you are considering and ask this plainly: if it is wrong, who notices first, who can stop it, and what record proves what happened?
“If AI touches patient communication, records, claims, imaging, scheduling, or clinical recommendations, the practice still owns the outcome.”
BY THE NUMBERS
300M+
people turn to ChatGPT with health-related questions each week, according to OpenAI
What it signals: Patients are not waiting for healthcare organizations to finish their AI committees. They are already bringing AI into how they understand symptoms, records, visits, and follow-up questions. Dental teams need a verification pattern before this becomes another awkward chairside moment.
Source: OpenAI, Health in ChatGPT, July 2026
READER Q&A
“What should a practice look for before trusting AI to become a habit instead of a novelty?”
Reader question
BH: The answer is already in the question. If you see AI as a novelty, it probably is not AI you should trust yet.
I think this is part of why so many dentists are still on the fence with AI imaging. I do not personally see imaging AI as a novelty. I think it can be an amazing sales tool and real support for the doctor in diagnosis, as long as it is not treated as a replacement for the doctor. But that is not the view across dentistry. What one dentist sees as trustworthy AI, another dentist may see as a gimmick, or even a threat.
So I would not start with the category. I would start with the practice.
What does your practice actually need? What would support real ROI, or create a genuinely better experience for the team and the patient?
A simple way to test this is to audit what you already spend money on. How much of it gets used every day? That is useful. How much of it sits on a shelf, or never gets logged in to? That is a novelty.
AI earns trust the same way. It has to become part of the work, not another thing the team has to remember to use.
FROM THE ORACORE BLOG THIS WEEK
Workflow | 2 min
Dental AI Workflow Agents Need More Than Scheduling Access
This is the closest OraCore piece to the agent story above. Scheduling access is not enough if the AI cannot carry context, respect approvals, and leave the team with a clean next step.
Evaluation | 4 min
Dental AI Weekly: Why AI Scribes Get Abandoned
This pairs with Brad’s Q&A. The useful test is not whether the demo impresses the room. It is whether the team still uses the workflow after the novelty fades and the schedule gets messy.
AI OPTIMISM
The useful AI story this week is control.
ChatGPT Health gives people a better way to organize questions and health context before they talk to a clinician. Presence starts with one job, limited access, policies, approvals, and escalation. The AI Act and the U.S. kill-switch proposal both point toward the same standard: if AI is going to act in important places, people need controls.
That is positive. The mature version of AI is not a black box making decisions in the background. It is assistive infrastructure: context when you need it, automation where the path is clear, escalation when judgment matters, and a record of what happened.
If you want to test AI where the workflow is simple, start with OraCore Scribe Solo or Team and make the note review habit real.
Try OraCore Scribe free for 14 days
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