Healthcare data and AI, written for the people who run health systems.
By Victor Wilson. I help health systems turn their data into money saved and operations that run better. New posts every couple of weeks on what works, what doesn't, and where AI actually fits.
-
What I would do in my first 90 days as a health system chief data officer
Listen to the people who own the money, settle a few disputed numbers, and land one real win before touching the org chart.
-
Interoperability after HTI-5 and a billion TEFCA records
More health records are moving between systems than ever, while federal AI transparency rules got lighter. Here's what to do about both.
-
The case for small, boring wins
Big data programs promise a lot and often stall. Small, plain fixes that save real money add up faster than most leaders expect.
-
Tariffs and drug shortages are a data problem now
Drug tariffs and rising shortages show up first in your purchasing data. Here's what to build so pharmacy and supply chain see them coming.
-
Your data team is too busy building dashboards nobody opens
Many health system data teams spend their weeks keeping alive reports that few people use. Here's how to find them and win that time back.
-
Four in ten hospitals are losing money. Here's what the data team can do this quarter.
The latest hospital margin data is grim. A data team can't fix it alone, but it can find real money before the year ends.
-
Where AI agents fit in health system operations right now
AI agents are ready for some back-office work today, with a person approving the output. Here's where I'd start and where I'd wait.
-
Denials are the revenue cycle problem data can actually move
Much of the revenue cycle is hard to change with data. Denials are different: the causes repeat and most of the fixes are in your hands.
-
How to run a 60-day pilot that actually decides something
Most pilots end in a shrug. Agree on the decision first, keep a control group, and let the result make the call.
-
Measure your analytics program in dollars, not dashboards
Count the money your analytics work saves or brings in, and stop counting the reports it ships.
-
Epic's agents are here: what stays your job
Epic showed its AI agents working together at UGM. They will do a lot inside the EHR. Three things still belong to your team.
-
Five questions to ask before any healthcare AI pilot
Many AI pilots in health systems never reach full use. Five plain questions, asked up front, tell you which ones will.
-
Your EHR is not your data strategy
The EHR's reporting database is a strong source. It is still only one source, and most of the questions leaders ask need several.
-
Fix the boring data before you buy the AI
AI tools are only as good as the data under them. The unglamorous fixes come first, and they pay off even if the AI never ships.
-
Medicaid work requirements: what your data team should build now
Medicaid work requirements start in January 2027, and coverage will churn. Here's the data a health system should start tracking now.
-
Supply chain is the cheapest margin a health system will ever find
Much of the savings in supply chain is money you already negotiated and aren't collecting. Finding it takes data work, not new contracts.
-
Price transparency files: what is actually useful in them
Hospital and payer price transparency files are messy and huge. A few narrow uses make them worth the effort.
-
Medicare's AI prior authorization pilot: what to track now
WISeR brings AI-assisted prior authorization to traditional Medicare. Early reports point to delays. Here is what to measure on your side.
-
Start with the CFO: why your first analytics wins should be financial
Your first analytics projects should save or recover money you can count. That is what earns the trust and budget for everything after.
-
Why most health system data platforms stall after year one
Year one of a data platform is about building it. Year two is about using it, and most plans never budget for that part.
-
Self-hosted open-weight models change the PHI math
Open-weight models are now good enough for many bounded healthcare tasks, and you can run them where your patient data already lives.
-
When provider AI meets payer AI, your data is the evidence
Providers and payers both use AI on claims now. The side with clean, traceable clinical data is the side that can prove its case.
-
MCP for healthcare agents: put the PHI rules in the server
Model Context Protocol gives AI agents one standard way to reach your systems. In healthcare, the access rules need to live in the server.
-
Treat vendor AI agents as a data source you don't control
EHR vendors are building AI agents into clinical workflows. Plan for their output as new data you need to label, test and watch.
-
Build healthcare AI that can survive a merger
Health systems merge and reorganize often. If your AI agents are wired to today's org chart, they won't make it through the next change.
-
Most of a healthcare AI agent should be ordinary code
In working agent systems, most steps are plain code. In healthcare, getting that balance right matters for safety, cost and audits.
-
What I learned building a data platform with AI agents
I built a working Snowflake data platform with AI agents in two days. The speed came from how we split the work.
-
Clinical notes are finally usable as data
Most clinical detail lives in notes, not codes. Language models have made it practical to turn that text into data you can query.
-
Getting FHIR data into Snowflake in a form people can use
Every health system has a FHIR API now. Here is an architecture for turning that data into tables your analysts can query.
-
The data engineer's job is changing
AI agents can now do much of the routine work of data engineering. The job that remains is design, context and judgment.