Physician. Behavioral Scientist.
Clinical AI Strategist.
I work with health systems, employers, technology companies, and physicians on a practical question: which parts of care can we safely entrust to AI? Together, we design the care model and work out how to test it.
The problem I’m interested in
Helping a physician get through the day faster is useful. But it leaves the same physician responsible for an ever-growing number of patients. Staffing shortages, long waits, and rising costs call for a closer look at how care is organized.
I’m interested in where AI can take on defined clinical tasks, with evidence that it works and a clear route back to a person when it doesn’t. That would give clinicians more room for the decisions and relationships that need them.
Who acts. Who reviews. Who is responsible.
AI gathers information, summarizes, and supports clinicians.
AI proposes clinical actions; clinicians approve.
AI executes bounded actions; clinicians review retrospectively.
AI operates within a defined scope; humans audit and manage exceptions.
AI independently manages validated workflows with defined escalation, surveillance, and accountability.
Progression depends on evidence, safeguards, monitoring, and clear accountability.
“The question isn’t whether AI can replace doctors. It’s which clinical decisions still require a doctor—and what evidence should be required before responsibility can safely move to AI.”
I help teams decide which parts of care AI can take on, which decisions need a clinician, and when the system should ask for help.
The work can include mapping a patient’s care journey, assigning responsibility at each step, and setting clear rules for escalation.
A model can perform well on a test and still fail in practice. I help teams design pilots that examine what happens to patients, clinicians, and the cost of care.
What gets missed? When do clinicians override the system? Who benefits, who is left out, and does the care actually improve? These questions belong in the evaluation from the start.
Will clinicians use the system? Will patients follow its advice? I bring behavioral science to the practical problem of knowing when to rely on AI and when to question it.
This means examining resistance, over-reliance, and the small decisions in a workflow or interface that change how people behave.
Identify which tasks AI could take on, where clinical judgment is needed, and when a person should step in.
Decide what evidence would show that it improves care, catches problems, and makes good use of clinicians’ time.
Identify a clinical workflow worth testing, given your patients, staff, technology, and constraints. Then define a manageable pilot, what to measure, and when to pause or stop.
An illustrative workflow · not a client case study
A connected cuff can send a reading. The harder work is deciding what happens next: what the patient hears, what the team sees, and who follows through. Explore one possible model below.
Fictional readings and scripted responses. This is a design example, not a working monitoring service or medical advice. No health data is collected here.
Routine check-in: Continue the agreed plan
01 / The next action
In this fictional case, the readings meet the care team’s agreed monitoring rules. The system records them and adds them to the next summary.
02 / Back to the clinical team
Trend summary with both source readings, measurement times, reported symptoms, and any missing information. No extra alert for this check-in.
03 / Responsibility
The assigned team reviews summaries on the schedule agreed before enrollment.
Measurement foundation: American Heart Association guidance on home monitoring. The collection and routing design above is illustrative.
Record readings, check completeness, provide approved measurement reminders, summarize trends, and route cases using clinician-approved rules. Any generated summary must be checked against the source data.
Diagnosis, medication changes, conflicting information, unresolved measurement problems, and cases outside the agreed population or protocol. The AI cannot rewrite its own escalation rules or expand its authority.
Before enrollment, specify eligibility and exclusions, response times, urgent instructions, out-of-hours cover, technical downtime procedures, and who owns each unresolved case. Assess the actual software functions and regulatory requirements before clinical use; this illustration is not a validated or cleared product.
The useful parallel is the operating boundary: which task, for which patient, with what data, and under whose supervision? A home-monitoring system should recognize when it has left those conditions and hand responsibility to a person.
The comparison has limits. Clinical care has different hazards and response times. A clinical inbox is not an immediate takeover mechanism. A safe handoff requires a named recipient, acknowledgment, a deadline, and a backup. Evidence from driving or aviation does not establish safety in healthcare.
The test is the whole care pathway
That is a question to test, not a benefit to assume. Compare with current care over a defined period, using the same outcome definitions and an appropriate study design.
Before expanding: test with synthetic failure cases; evaluate in shadow mode without changing care; then run a limited, supervised pilot with predefined success measures and stop rules. Reassess after material model or workflow changes.
The Future of Care
I’m developing a series of conversations with people who see healthcare from different angles: clinicians, technologists, employers, payers, and colleagues in behavioral science, policy, cybersecurity, and finance. I want to put difficult questions on the table before the answers are treated as settled.
Join the ConversationOn the agenda
Which of our clinical workflows could AI manage safely—and which need a human in the loop?
How do we test AI-enabled clinical workflows?
Can AI make healthcare cheaper—and care more reliable?
Conversations coming soon.About Angela
Angela Thyer, MD, MSc
Physician · 20+ years of clinical experience
After more than two decades practicing medicine, I became increasingly interested in a problem medicine alone couldn’t solve: knowing what patients should do is very different from helping people actually do it.
That led me to behavioral science. AI now raises a second question: if machines can increasingly perform parts of clinical reasoning and care delivery, how should healthcare be redesigned around that capability?
These are the questions I now work on with organizations planning how to use AI in care.
Let’s Talk
If you’re working through a care model, planning an AI pilot, or have a question you’d like to discuss, I’d like to hear from you.