No Staff, No Problem? How Independent Physicians Are Using AI to Run Their Practices

Offcall and Ultralight hosted a free workshop on what an AI-supported independent practice actually looks like day to day.

Dr. Graham Walker moderated, joined by Dr. Michael Hobbs, a direct primary care pediatrician who runs a solo practice with no staff, and Dr. Hillary Lin, co-founder and CEO of Curator, who runs a cash-pay longevity clinic, a health tech company, and a media business with a single virtual assistant. Sunita Mohanty, founder and CEO of Ultralight, presented the company's AI-native EHR, followed by a live demo from a physician who runs his cash-pay virtual primary care practice on it.

The session stayed practical. Rather than predicting where AI is headed, the speakers walked through the specific letters, forms, websites, and workflows they have already handed off, the rules they follow to keep patients and data safe, and how they think about building their own tools versus buying them.

Key Webinar Moments

• Michael Hobbs explains how he runs a solo pediatrician with zero staff
• Michael suggests you pick one annoyance and spend fifteen minutes on it
• The rule every speaker agreed on: let AI draft, never let it send
• Why Dr. Hillary Lin says the problem isn't patients using AI, it's patients who don't know how to challenge it
• A live demo of an AI-native EHR producing a visit prep, a clinical synthesis, and a patient-facing 90-day plan
• Graham Walker's advice for the physician who feels intimidated: start by scheduling a dinner on your personal calendar

Start With One Annoyance, Not a Grand Plan

Michael opened by describing the reality of ownership: every problem and every success lands on your desk, from scheduling to IT to hiring. His advice was to find the friction point you wish someone else handled, whether that is the same forms you send every new inquiry, patient letters, handouts, or prior authorization drafts, and describe it to ChatGPT or Claude in plain English. Then tell it what it got right and wrong, and iterate. He framed AI tools as fractional employees: give each a small task, review the work, give feedback, and expand its responsibilities only as it earns trust. Expect to reach roughly 75 to 80 percent and keep tweaking from there.

Graham added that once a workflow does what you want, you can ask the tool to turn it into a reusable skill so you do not have to rebuild it next time. Michael described his own advanced setup, an agent that connects his communications platform to his EHR and can book appointments, add an allergy, or pull information on request, and shared that one patient told him the AI triage replies sounded just like him. Both cautioned that these tools sometimes fill in details you did not ask for or quietly edit out details you wanted, so review every draft.

Draft, Don't Send

Michael and Graham landed on the same guardrails. Make sure any vendor touching protected health information has signed a business associate agreement, or keep PHI out of the tool entirely. Let AI draft communications to patients, vendors, and peers, but keep a human on the send button. Give the tool the least access it needs, and keep a record of what it does so you can step back if something goes wrong. Graham tells every email-connected agent it is never allowed to send, and suggested practicing with low-stakes personal tasks, like a family calendar, to learn how these tools fail before patient data is anywhere near them.

Michael also noted that employed physicians generally cannot sign a BAA on their own, because it has to happen at the organizational level. On AI avatars and cloned voices, he and Graham agreed it comes down to where you personally draw the line on authenticity with patients.

Running a Practice Through an AI Harness

Hillary described how she runs her patient flow, website, booking, payments, and content through what she calls a harness: software that deploys AI agents across your tools and consolidates your tasks in one window. She builds on an off-the-shelf harness rather than custom code, reads every message and reply herself, and uses the AI to prep each next step, whether an email or a strategy, which she then verifies and edits. She said scheduled tasks were a game changer, such as a daily run through her calendar and email that tells her what needs attention, and Michael reminded the audience to test any scheduled task before relying on it.

Hillary also walked through the two major platforms. She reported that OpenAI offers more routes for a solo practitioner to get a BAA signed, that Anthropic's protections generally require an enterprise plan, and that ChatGPT for Clinicians is free with an NPI. Those details change quickly, so confirm current terms with each vendor.

What HIPAA Requires, According to Dr. Lin

Hillary shared her reading of what a small practice actually needs: a BAA with any vendor that touches PHI, a written risk analysis, which AI can help draft, and the ability to give patients their records within 30 days. In her view, HIPAA does not require a specific certified EHR or a single vendor, and some cash-pay practices may fall outside HIPAA unless they conduct insurance-related transactions such as prior authorizations. She stressed this is one clinician's interpretation. Check your own situation with counsel or a compliance resource.

She also pointed out that many standalone tools are becoming features of the large AI platforms, but that purpose-built tools still earn their place, particularly for patients, and that building everything yourself can get complicated without time or help.

Teaching Patients to Challenge Their AI

Many physicians are frustrated when patients arrive with an AI-generated plan. Hillary's view is that the issue is not that patients use AI, but that they do not know how to question it. Her suggested prompts are simple: ask it which questions to bring to your doctor, and ask what it does not know about you that would change its answer. Michael added that models tend to want to please the user, so both clinicians and patients need to learn to prompt against that tendency. He also raised a privacy point to share with patients: records uploaded to a general consumer chatbot may not carry the protections of a purpose-built tool.

An AI-Native EHR, Demoed Live

Sunita described Ultralight as an EHR built as a context engine for independent and direct care practices, where records, labs, wearables, and patient messages are all available to the AI, and clinicians can run their own SOPs against each chart. She said the platform is live in more than 100 clinics, with 30 added last quarter, and that clinicians tell her it helps them catch things in old records they would otherwise have missed.

The demo physician, an internist who recently opened his own cash-pay virtual practice, showed three uses on a fictional patient. First, a visit prep prompt that summarizes the prior encounter, interval labs, wearable data, open loops, and a suggested agenda. Second, a clinical reasoning prompt that produces a synthesis, red flags, a differential, a workup strategy, and cited evidence, through whichever framework the physician prefers. Third, a patient-facing executive health report ending in a 90-day action plan. He noted he generated the prompts by describing what he wanted to the copilot in plain English. As with any sponsor demo, ask vendors for specifics about validation and data handling before you commit.

The session closed with a bet from Sunita: in a landscape that feels increasingly faceless and branded, physicians who practice with a clear philosophy and a real relationship with patients may have an edge, and AI makes running that kind of business more feasible than ever. The webinar also pointed attendees to the Offcall app and its physician-only independent practice lounge, where members can swap workflows and ask questions of colleagues.

Resources and Links

• Michael Hobbs' getting-started guide: ai101.health
• Ultralight: https://www.linkedin.com/company/ultralighthealth/
• Hillary Lin, Curator: https://www.linkedin.com/company/curatormed/
• Dr. Lin's slides and summary: https://curator.med/resources/offcall
• Join the Offcall independent practice physician lounge in the Offcall app

This webinar was sponsored by Ultralight. Dr. Hobbs and Dr. Lin are advisors to Ultralight.

Follow Offcall on LinkedIn to be notified about upcoming sessions.