Clinical AI & Innovation

AI is a workflow decision before it’s a technology decision.

Health systems don’t need more pilots. They need pilots designed to answer a question, measured honestly, and built to scale. Here is what that looks like in practice.

Case study 01

Pilot → Scale

Cleo AI ambient scribe · USACS Hospital Medicine · Carson Tahoe Regional Medical Center (Epic)

Taking an AI scribe from feasibility to enterprise scaling

The pilot showed that workflow redesign and prompt optimization, not the software alone, drove the gains in documentation quality.

The challenge

Ambient AI scribes promise relief from documentation burden. But hospital medicine notes carry quality, coding, and documentation-integrity requirements that a generic tool can miss. Leadership needed evidence, not enthusiasm.

The approach

I designed a before-and-after evaluation with measures defined in advance: a multidomain clinical documentation integrity (CDI) quality scorecard and NASA-TLX–adapted physician workload surveys. I then led implementation, outcome measurement, and physician onboarding.

Before & afterMultidomain CDI quality scorecard
NASA-TLXAdapted physician workload surveys
1 of 5Voting physicians on enterprise templates

The result

Documentation quality scores improved across domains, with the largest gains after prompt optimization. USACS has since moved the program from feasibility to enterprise scaling. I now serve as one of five voting physicians setting enterprise H&P and progress-note standards, clinical decision nudges, and CDI-aligned rules for the AI scribe.

Presented at InnovatorMD

“AI Medical Scribes in Hospital Medicine: From Technology Adoption to Clinical Transformation.” Request this talk

Five questions every board should ask about an AI pilot

What does good look like before go-live? Who owns the measures? What will we standardize for every clinician, and what stays customizable? How will we know whether the workflow or the tool drove the result? What evidence would make us stop?

Case study 02

35%

Fewer sepsis bundle fallouts · Hospital medicine, Arkansas (Meditech)

Turning protocol design into measurable quality and cost gains

A protocol-driven sepsis alert and revised COPD order sets reduced sepsis bundle fallouts by 35% and levalbuterol use by 34%.

35%Fewer sepsis bundle fallouts
34%Less levalbuterol use
≈ $20KEstimated annual savings

Why it matters

Quality and cost aren’t competing priorities when the work is designed well. Standardizing the right decision at the right moment improved bundle compliance and reduced waste at the same time.

Leadership role

As Chair of Medicine and Chief Hospitalist, I led the hospitalist program, quality, and hospital-physician alignment. I served on the Quality Improvement, Pharmacy & Therapeutics, Antimicrobial Stewardship, and Utilization Management committees and on the executive order-set committee.

Case study 03

Access

Direct-admission protocol · Featured in Today’s Hospitalist

A direct-admission pathway that strengthened referral relationships

Clear guidance for primary care physicians on hospital admission versus ED referral.

Referral growth depends on how easy a hospital is to work with. The protocol replaced an inconsistent, informal process with a predictable pathway, and it strengthened the relationship between community primary care physicians and the hospitalist team.

Case study 04

3 products

Doximity Dialer, Doximity Scribe, and Doximity Ask

Physician feedback that shaped digital health products

Named Doximity Digital Health Advisor of the Year (2024) for physician feedback that shaped three of its products.

Clinical products succeed when they are built with physicians, not just for them. Through structured testing and feedback, I help product teams see the workflow realities that decide whether a tool gets adopted or abandoned.

I bring the same lens to QuickBlox, contributing clinical design input on agentic AI triage and telemedicine workflows, and earlier to MarianaAI on AI scribe product feedback and regional health system introductions.

Virtual care at scale

Multi-state virtual care, from the inside.

29 state medical licenses plus the Interstate Medical Licensure Compact, with APP supervision across several telehealth platforms.

Kandu Health

Post-stroke neurostimulation

Principal Illness Navigator and Ipsihand prescriber, one of a small number of physicians credentialed for this device.

Oscar Health · OpenLoop Health

APP supervision

Virtual urgent care and supervision of advanced practice providers across a multi-state model.

Access TeleCare

Telehospitalist coverage

Remote hospitalist coverage for hospitals in Nevada, Texas, and Kansas.

Start a conversation

Building a board, a clinical AI strategy, or a medical leadership team?

I welcome conversations about paid, nonprofit, and advisory board service, fractional CMO roles, clinical AI advisory, and speaking.