AI for Medical Practices
REAL AGENTIC AI ADOPTION IN HEALTHCARE
AI FOR MEDICAL PRACTICES
Every software pitch a practice owner hears now includes AI. The EMR drafts notes, the phone system answers calls, and the marketing platform writes the posts. AI for medical practices, done well, goes further: AI agents that carry out multi-step work across the practice's own data and workflows, under rules the practice sets, with a clinician or manager reviewing what they produce. That is where the gains in efficiency, productivity, and consistency of care come from.
​
Luminary helps owners of medical aesthetics and wellness practices decide where AI belongs in the practice, how to adopt it, and how to govern it. Luminary is paid only by the practice. It sells no software and takes no fee or commission from any AI company, software vendor, or developer. Its advice on what to build or buy has no stake in the answer.
TWO KINDS OF AI A PRACTICE OWNER IS OFFERED
Most of what is sold to practices as AI is a feature inside software the practice already rents. These features can be useful, and a practice should use the ones it already pays for. Each one works inside a single product, on the data that product holds, and on that vendor's roadmap.
​​
The second kind is an agentic system built around the practice itself. It reaches across the practice's systems, follows the practice's protocols, and carries a task from start to finish, stopping at the points where a person has to review or approve. The two are easy to confuse because both are called AI. They differ in what they can reach, who directs them, and what the practice keeps.
​
WHAT AGENTIC AI MEANS IN A MEDICAL PRACTICE
Generative AI produces something when asked: a draft, a summary, an answer. Agentic AI works toward a goal. It breaks the goal into steps, uses the tools and data it has been given, checks its work against rules, and hands the result to a person at a defined point. A generative tool can draft a follow-up email. An agent can find every patient whose treatment plan calls for a follow-up this week, pull what each one received and when, draft the message that fits each patient, and place the drafts in front of the coordinator for approval.
​
For an owner, the distinction matters for three reasons. Agentic systems do whole pieces of work, so the time they save is counted in completed tasks. They run on the practice's own data and rules, so the output reflects how the practice actually operates. And because they are designed around the practice, the review points, the access limits, and the patient disclosures are decisions the owner makes deliberately.
​
WHERE AGENTIC AI CHANGES A PRACTICE
The best candidates are workflows that are repetitive, rule-based, and spread across several systems, where staff spend their time assembling information before they can use their judgment. Each example below names the point where a person reviews the work.
-
Intake and preparation. An agent assembles the new patient's forms, history, and consult notes into one summary before the visit. The provider reviews it before the consult.
-
Treatment planning drafts. An agent drafts an individualized plan from the consult, the patient's history, and the practice's protocols. The treating clinician reviews, edits, and signs it.
-
Follow-up and recall. An agent reads treatment histories to find patients who are due, overdue, or missing a step in their plan, and drafts the outreach. Staff approve every message before it is sent.
-
Financial and KPI reporting. An agent pulls the month's numbers from the practice's own systems and shows the source of each figure. The owner or manager checks the report against the books.
-
Documentation quality. An agent checks charts against the practice's documentation standards and flags gaps. The clinician makes the corrections.
-
Protocols and staff questions. An agent answers staff questions from the practice's own protocols and policies and cites the document it used. A manager owns the protocols it draws from.
​
Quality of care improves through consistency: every plan built from the same protocols, every follow-up sent on time, every chart checked. Clinical decisions stay with licensed clinicians. The same principle behind structure before scale applies here: a workflow has to be sound before it is automated, because an agent will repeat a broken process as faithfully as a good one.
​
HOW TO USE AI IN YOUR MEDICAL PRACTICE: SEVEN STEPS
-
Map the workflows and the data first. List where staff time goes each week and which systems hold the information each task needs. The map shows where AI could help before any product is chosen.
-
Use the AI features you already pay for. Many practices own AI features in their EMR, phone, or marketing software that are switched off or unused. Turn on the ones that fit a mapped workflow.
-
Pick one workflow with a measurable result. Choose a task where the before and after can be counted: hours per week, days to follow-up, charts completed, plans delivered.
-
Decide whether to build, buy, or configure. Some workflows are best served by an existing product. Others need an agent built on the practice's own data. Decide on the workflow's merits, with an advisor who has no stake in the answer.
-
Settle the data questions. Sign a business associate agreement with any vendor that will touch patient information, confirm where the data is stored, and limit access to what each task needs.
-
Set the review point and the patient disclosure. Name the person who reviews each output before it reaches a patient or a record, and decide how patients will be told that AI is used in their care.
-
Measure against the baseline, then expand. Compare the result with the numbers from step 3. Extend the approach to the next workflow only when the first one is working.
​
AI FOR MED SPAS AND WELLNESS PRACTICES
Cash-pay aesthetics and wellness practices have their own pressure points: consult-to-treatment conversion, membership and package follow-through, recall on treatments that need maintenance, provider utilization, and protocol consistency across injectors and staff. Each of these depends on information that sits in several places at once. That is the kind of work agentic systems handle well.
​
Most published guidance on AI for med spas is a list of tools: an AI receptionist, a chatbot, a content generator. Any of those may be the right purchase for a given practice. The larger decision sits above the tool list: which workflows matter most to this practice's economics and care, what the practice already owns, and what should be built around its own data. That decision is where Luminary works.
​
TEXAS AND FEDERAL RULES FOR AI IN A MEDICAL PRACTICE
1. Patient disclosure
(Texas HB 149, effective 1/1/26)
When an AI system is used in relation to a health care service or treatment, the provider must disclose it to the patient no later than the date the service is first provided, or as soon as reasonably possible in an emergency. The disclosure must be clear, conspicuous, and in plain language, and it is required even when the AI use would be obvious. The statute leaves open exactly which tools count and what form the notice should take, so the wording belongs with your attorney.
​
2. Practitioner review & disclosure
(Texas SB 1188, effective 9/1/25)
A practitioner may use AI for diagnostic purposes, including recommendations on a diagnosis or course of treatment, when acting within the scope of their license and reviewing all AI-created records under standards set by the Texas Medical Board. The practitioner must disclose that use to patients.
​
3. Where records are stored
(Texas SB 1188, from 1/1/26)
Electronic health records must be physically maintained in the United States, including records held by a cloud, software, or AI vendor on the practice's behalf. Any AI system that stores or processes record data needs to meet this requirement.
​
4. Business associate agreements (HIPAA)
Any vendor that creates, receives, stores, or transmits patient information for the practice is a business associate and needs a signed business associate agreement before any patient information reaches it. That includes AI tools.
​
4. Business associate agreements (HIPAA)
Practices that receive federal funds, such as Medicare or Medicaid payments, must make reasonable efforts to identify and reduce discrimination risk in tools that support patient care decisions, including AI tools. Most cash-pay med spas fall outside this rule; practices that bill Medicare should confirm their status with counsel.​
​
HOW LUMINARY HELPS PRACTICES ADOPT AI
Luminary's principals run agentic AI systems inside their own medical aesthetics and wellness practices. One drafts individualized treatment plans from consult notes and the practice's protocols for the treating clinician to review and sign. Another reads the practice's own treatment and financial data to find patients due for follow-up and services that commonly belong alongside the care they already receive. The advice comes from owners who have designed, governed, and paid for these systems themselves.
​
Luminary advises on AI adoption as part of an Advisory Engagement on operations and scaling. The engagement maps the practice's workflows and data, identifies the AI features the practice already owns, chooses the first workflow, and makes the build, buy, or configure decision. It also evaluates any vendor or developer proposal on its merits and sets the governance: business associate agreements, data storage, review points, and patient disclosure. You receive a written framework and a sequenced roadmap, including the measures that will show whether it is working. The build is done by the practice's own team or a developer the practice chooses, and Luminary can review the finished system against the design.
​
Owners who are not yet sure where the practice's real constraints are start with a Practice Diagnostic, because AI pays off only on a workflow that is already sound. Owners opening a practice can raise AI in Launch Advisory, while the workflows are being designed for the first time.
​
Luminary does not write code, configure software, host systems, or resell any product, and it takes no fee from any AI company, software vendor, or developer.
​
