1. 01

   Start with the workflow, not the model.

   Map one process end to end: inputs, handoffs, people, systems and the step where it breaks. AI goes where a handoff already fails.

   From the record At Optum Advisory Services I translated clinical requirements into technical workflows on Epic EHR deployments.

2. 02

   Decide what right looks like before building.

   Score the task checkpoint by checkpoint instead of one pass or fail, and agree in advance what counts as a miss.

   From the record I architect multi-step scoring rubrics for 15+ AI evaluations.

3. 03

   Tie it to a line someone already tracks.

   Revenue, reimbursement, cost or time to the next care step. A change that moves none of them will not survive the budget cycle.

   From the record I modeled CMS-HCC V24-to-V28 reimbursement shifts for 25+ clients.

4. 04

   Keep a human owner on every handoff, and pilot where it cannot hurt anyone.

   De-identified or fictional data first, and PHI terms agreed in writing before anything else.

   From the record ChartAuth runs on fictional records only.

• UnitedHealth Group, Optum Advisory Services ·2022 to 2024

  A reimbursement shift, modeled for 25+ clients

  Result5x ROI in Year 2

  Situation

  Clients faced the CMS-HCC V24-to-V28 reimbursement shift.

  What I did

  As a Senior Consultant at Optum Advisory Services, I modeled the V24-to-V28 reimbursement shifts for 25+ clients using historical data.

  Takeaway

  In healthcare the business case runs through reimbursement. When I scope AI work I start in the same place: the paid step, the denial or the cost line the change would touch.

• UnitedHealth Group, Optum Advisory Services ·2022 to 2024

  Clinical requirements, turned into technical workflows

  Result25% fewer 30-day hospital readmissions

  Situation

  Backend Epic EHR product updates had to fit the way care is delivered, and the engineering teams building them were offshore.

  What I did

  I managed 10+ offshore engineering teams deploying backend Epic EHR product updates, and translated clinical requirements into technical workflows supporting care delivery.

  Takeaway

  The hard part was the translation between the clinician and the system. Model output has the same problem: someone has to turn what the clinician means into something the system can check.

• Dioun Health ·2025 to 2026

  An EMR standardization inside a bolt-on acquisition

  Result$800K in annualized operational cost synergies

  Situation

  A $15M bolt-on acquisition for an existing home health portfolio company, with specialized home health EMR software to standardize as part of the integration.

  What I did

  As a Private Equity Analyst in the Healthcare Acquisitions Group, I managed end-to-end execution of the acquisition, including standardizing the specialized home health EMR software.

  Takeaway

  Technology value arrives when the workflow is standardized, not when the software is bought. I expect the same of AI: the value shows up when the work around the tool changes.

Two fixed-scope ways to start, each priced up front.

• Diagnostic session$500One 90-minute working session, then a written plan within two business days.
• Working sprint$1,000A half day of hands-on work, then a 30-minute follow-up call within the week.

• WagerThesisApplied AI Markets·LaunchingExplore live markets, understand the price, and build and test your thesis with simulated funds.wagerthesis.ai(opens in a new tab)
• FindYourHumanHuman marketplace·LiveHire a human. Or be one. Find human expertise for AI work and approved, low-risk local tasks.findyourhuman.ai(opens in a new tab)