This website uses cookies

Read our Privacy policy and Terms of use for more information.

On October 30, Offcall and Ultralight hosted a live workshop on how independent physicians use AI to run their practices. Dr. Graham Walker, Offcall's co-founder and an emergency physician, moderated. Dr. Michael Hobbs, Dr. Barry Breaux, and Dr. Hillary Lin walked through their own setups, and Sunita presented Ultralight.

Dr. Hobbs opened with a shared experience most readers live with daily. When you own the practice, he said, "everything at the end of the day falls in your lap, every problem, every issue, but also every success." He runs a solo direct primary care practice with no staff, so he books visits and fills out forms himself. Dr. Lin runs her longevity practice with one virtual human assistant, who she said "does less and less every day as I have AI do more and more."

The panel had different ideas about how quickly to integrate AI into clinical practice. Dr. Lin encouraged the audience to "dare to dream" and shared how she builds most of her own tools, right down to the booking flow on her website. Drs. Hobbs and Walker were more cautious: start small and be clear about what you'll let AI do. Dr. Walker's first rule for any tool connected to his email is one I share: nothing gets sent without his review. Still, three common threads emerged from the panel that apply to any practice.

1. Hire AI the way you hire staff

Dr. Hobbs called these tools "fractional employees." Write the job description, give a small first task, check the work, and widen the role once it earns trust. His rules were concrete: a business associate agreement with any vendor that touches patient data, drafts only, and "the least access to your information" the job needs. He also warned that a model "will fill in gaps that you don't necessarily want it to fill in."

For a clinician who is also the front desk, this changes how you spend your pajama time. Dr. Hobbs also mentioned that most first projects get 75 to 80 percent of the way there on the first pass with AI, and that is already useful. When he trained an agent to handle basic triage replies, the feedback he valued most actually came from a patient: "It sounds like you, Dr. Hobbs."

2. Choose tools you can trust with patient care

Dr. Lin described how she works with several HIPAA-compliant vendors, with a BAA in place for each one that handles patient data. She was also candid about the work involved. Building your own clinical system takes time most clinicians don't have, and without that investment, "it can get rather hairy." Sunita explained why those details matter: generative AI "can still make mistakes," including misreading a document. When that information becomes part of a patient's record, you need a way to check what comes in and what gets carried forward.

For me, that shapes what you build. A website is a reasonable place to start: you can review it, catch mistakes, and make edits before it goes live. Inbox triage needs more care, especially when it involves patient information or clinical questions. The chart still needs closer attention, because an error there can shape a care decision. Dr. Lin maintains hers as a linked set of documents. Dr. Breaux uses Ultralight, where the chart updates as new labs and messages arrive. Either approach leaves the clinician with the same responsibility: knowing how information gets into the record and checking that it's accurate.

3. Provide continuous care that fits in a day's work

Dr. Lin described her care plan as "alive, living, and continuous," with messages going back and forth several times a day. She credited AI with making that level of personal care possible. In a membership practice, that ongoing contact is part of the promise. It's also what makes the work hard to sustain.

Dr. Breaux showed how he gets help with it. One prompt brings the prior plan, open items, lab trends, wearable data, and recent messages into a prep sheet he keeps open during video visits. Another produces a clinical synthesis using a conventional differential and a functional medicine lens, with citations he can check. A third creates a patient report with a 90-day action plan. He built all three with the copilot by describing what he wanted in plain English.

What stayed with me was what it found in old records. Dr. Breaux said AI keeps surfacing buried findings, including an aneurysm on a CT scan from years earlier that the patient hadn't known about. He verifies each finding himself. Dr. Walker described reading 53 pages of nursing facility records to find the one page he needs. He called it a human memory shortage. That's a familiar constraint when you're trying to follow patients over time: the information may be there, but finding it and keeping it in view takes work.

Nobody on the panel has outcome data yet. We have three practices' experiences and a live demo. Sunita reported that Ultralight is live in more than 100 clinics and handled more than 12,000 clinical questions in October. That tells us clinicians are using it. Whether it improves care is still an open question. But hearing Dr. Breaux say "medicine has become fun again" stayed with me. If AI can make the daily work more manageable and give clinicians more room to think, that's worth paying attention to while we wait for the outcome data.

What to try this week

  1. Pick one friction point in your workflow and work with AI to create a solution. Dr. Hobbs suggests a letter you write often, a handout, or the reply to a common inquiry. Describe the problem candidly, tell the AI what it got right and wrong, and stop at good enough.

  2. Save the version you like. Once the output is right, ask the tool to turn it into a reusable skill or template so you can call it again.

  3. Write your house rules before you connect anything. Put a BAA in place before any patient data, use drafts only, and grant only the access the task needs. Practice on your personal calendar or email first.

  4. Run a visit-prep prompt on your next three patients. In a HIPAA-compliant clinical tool, try something like: "Prep this visit. Summarize who this patient is, what we decided last time and what is still open, new labs and trends since then, anything in recent messages I should know, and a suggested agenda for today."

  5. Ask what the old records hold. Have your clinical AI list findings in outside records that never got a follow-up, then verify each one yourself before you raise it with the patient.

Be a modern clinician with Ultralight, the AI-native EHR built specifically for functional, integrative, and longevity medicine. We've recently launched wearables integrations and improved AI-native clinical workflows. Get in touch to see the latest updates.

Watch: How Independent Physicians Are Using AI to Run Their Practices

The full October 30 workshop with Dr. Graham Walker, emergency physician and co-founder of Offcall (LinkedIn); Dr. Michael Hobbs, pediatrician, AI educator, and founder of Lakes Pediatrics; Dr. Hillary Lin, Stanford-trained internist and co-founder and CEO of Curator (site); Dr. Barry Breaux, internist and founder of Breaux Medical; and Sunita Mohanty, Founder and CEO of Ultralight (LinkedIn). It includes Dr. Breaux's live demo.

In the news

Modern hormone therapy showed no added heart risk in women who started it between 45 and 59. In a study presented at the American Heart Association's Scientific Sessions this month, researchers followed more than 52,000 women in Israel for an average of 12 years. Starting contemporary hormone therapy, in oral, patch, or gel form, was not associated with a higher risk of heart attack, stroke, blood clots, or death. It is an observational research abstract and has not yet been published in a peer-reviewed journal. For those of us who prescribe, it adds reassurance for the patient who hesitates because of heart risk, and it fits the current guidance to start within 10 years of menopause onset and before age 60.

Patients are comfortable with AI in their care when a clinician can overrule it. A West Health-Gallup survey of 5,330 U.S. adults who had seen a primary care provider in the past year found 56% were comfortable with AI making diagnosis and treatment decisions that a provider could override. With no provider review, that fell to 9%. A companion survey of 1,007 primary care physicians found only 36% expected most patients to be comfortable with the override arrangement. Both surveys asked about hypothetical arrangements and did not measure outcomes. When we tell patients how we use AI, the key point is that we review every recommendation before it reaches them.

A large AI scribe study found less charting time, longer notes, and no productivity gain. Published in late October in Annals of Internal Medicine and worth the late look, the quality improvement study followed outpatient clinicians across a 14-hospital Boston health system after an ambient scribe rollout. Median note-active time fell from 21 to 18.6 minutes, and median note length rose. Same-day and after-hours note closure did not change, and neither did clinical productivity. A scribe shortens the note. The rest of the evening still goes to chart review, messages, and follow-up, which is where the next real time savings sit.

Upcoming Conferences & Events

Dec 11-13, A4M Longevity Fest · Las Vegas, NV · The biggest longevity event in the U.S. The room spans clinicians, industry, founders, and the people building next year's platforms, and the connections from this one tend to compound through the rest of your year. Save $100 on registration with code Longevity26. OvationLab and Ultralight will be there!

Feb 19-20, CALM 2027 · Fort Lauderdale, FL · NuBioAge's Clinical Application of Longevity Medicine meeting for physicians, NPs, pharmacists, and clinical teams putting longevity medicine into practice. Ultralight will be there!

Know of an event we should add? Reply and tell us.

Until next week

The physicians on this panel handed AI one job at a time and kept the chart in a system built to hold it. Pick one friction point this week and start there.

Reply and tell us the first task you would hand to AI in your practice. Know a colleague running a practice mostly on their own? Forward this to them.

Until next week, keep building the practice you imagined when you started.

Sunita & Dr. G