Billy·Oglesby

Hospital-system strategy simulation

The problem

Healthcare strategy is taught from lectures and cases, which means students analyze decisions other people already made. What they rarely get is the experience that actually teaches strategy: running a hospital system against live competitors, with imperfect information, advisors who disagree, and consequences that compound round after round. Commercial simulations exist, but none were built for how AI changes the advisory layer itself.

The design

A multiplayer strategy simulation, built as a companion to the Modern Healthcare Strategy textbook. Teams of students govern competing hospital systems in one shared regional market. Each round, a team consults its LLM division directors (AI advisors standing in for the executives a real board hears from), then submits a decision memo. A deterministic engine resolves the round and returns reports.

One round of the hospital-system simulation A cycle: student teams consult LLM division directors, decide and submit a round memo, a deterministic engine resolves the round, and reports return to the teams for the next round. The human decision is highlighted; the AI advises and the engine computes. Student team decides governs its hospital system · submits the round memo Deterministic engine resolves the round · versioned per cohort no model computes a number Round reports market, finances, and outcomes, back to every team LLM division directors advise from their division’s view fixed tool contracts · scoped information memo in results out next round advice The AI advises. The humans decide. The engine keeps score.

The division of labor is the point, and it is the same line I hold everywhere else:

  • The AI advises. Division directors argue from their division's view of the market, through fixed tool contracts with validators. They can read the state they are entitled to see and nothing more.
  • The humans decide. The round memo is the team's judgment, not the model's output, and it is graded as such.
  • The engine keeps score. Round outcomes come from a pure, deterministic, versioned numerical engine. No model ever computes a financial result, so identical decisions always produce identical outcomes, a fairness property you cannot get from a model and a prompt.
  • Information boundaries are enforced. Teams see their own system and the public market, never a competitor's internals.

The numbers

Figure Value
Pilot January 2027
First cohort 4 teams × 5 students
Round outcomes computed by AI 0 (by design)
Status in build; pilot site at simulation.modernhealthcarestrategy.com

What went wrong and what I changed

Nothing yet, and I do not trust that. The build is pre-pilot; the first incidents and fixes will be logged here, dated, from the January 2027 cohort.

What I’d do next

Run the January 2027 pilot cohort and publish what the teams did that the design didn't predict.