Five products and the lessons I still use
Each product turned on who would carry the work: ownership after a hospital install, approval for a physician’s phone line, a frightened family’s attention, accountability for a nurse’s queue, and who pays.
Five deciding case studies from 2012 to 2024. For active systems and ventures, see current work.
| Value | Measure | Source |
|---|---|---|
| 100+ | organizations in federal quality reporting | Epic — the quality-reporting curriculum and escalation path I ran for its clients |
| 300,000+ | calls on an average workday | Doximity Dialer, which I originated — Doximity’s figure, February 2026 |
| 30 million | annual visitors | CancerCompass — I directed digital products |
Case studies
| Study | Decided by | Outcome | Years |
|---|---|---|---|
| Go-live is where the job starts | Ownership | An account at risk of cancelling became a reference account, a 134-year-old health system took its MyChart patient portal and EpicCare Link live, and all eight client organizations were still reporting on the federal quality programs when I left. | 2012–2013 |
| The patient sees the office number, not the doctor’s cell | Approval | Dialer still carries more than 300,000 calls on an average workday, 9 years after I left. | 2013–2017 |
| Plain-language next steps for 30 million annual visitors | Attention | Chat conversions up 267% after the rebuild I led. | 2019–2021 |
| Four care programs, one system, and an owner for every case | Accountability | Four care programs launched on one shared system, and the gap between the best and worst service lines narrowed. | 2021–2022 |
| Reading a physician’s employment contract before they signed it | Who pays | 1,200+ physician users; sustained engagement and funding fell short. I wound it down in 2024. | 2022–2024 |
-
Go-live is where the job starts
An account at risk of cancelling became a reference account, a 134-year-old health system took its MyChart patient portal and EpicCare Link live, and all eight client organizations were still reporting on the federal quality programs when I left.
What I’d keep Name the trigger, the owner, and the notification before the problem arrives.
MyChart as Epic presents it today. The version I took live for the research hospital was the 2012–13 one. -
The patient sees the office number, not the doctor’s cell
Dialer still carries more than 300,000 calls on an average workday, 9 years after I left.
What I’d keep If the person choosing carries the risk, adoption can come before official approval. If the risk is clinical, it can’t.
The identity decision in the product’s own words: the clinician picks which number the patient sees, with the office line preselected. -
Plain-language next steps for 30 million annual visitors
Chat conversions up 267% after the rebuild I led.
What I’d keep A frightened person is not short of information. The scarce resource is attention and trust, so the product’s job is the next step and a person to call.
The homepage in July 2019, as the Wayback Machine archived it: the article-and-message-board site my rebuild replaced. -
Four care programs, one system, and an owner for every case
Four care programs launched on one shared system, and the gap between the best and worst service lines narrowed.
What I’d keep One system can carry many programs once every decision point has a named owner.
The shared decision architecture, drawn from the case study rather than from a surviving document: four programs with four sets of protocols enter one routing layer, and the layer assigns each case to a clinical role. The alternative on the table was four products, one per lane. The record beneath is what let a nurse and a navigator see the same case. -
Reading a physician’s employment contract before they signed it
1,200+ physician users; sustained engagement and funding fell short. I wound it down in 2024.
What I’d keep Who pays decides who the product works for. Sign-ups show people want it; they do not settle that.
The Contract Analyzer as specified in January 2023. This is the design artifact, not a screenshot of a live analysis — the document body is placeholder text.
Where the rules go next
Two of them carry straight into what I build now. Ambit applies the approval rule to AI agents, whose risk is not the user’s alone: an agent gets each new action only once someone approves it. NextConsensus applies the ownership rule to medical evidence: it keeps watching a guideline after publication, to catch the evidence moving past it.
Revised