My role at Hybridge is titled AI engineer, but the actual work is closer to embedded engineering: the stakeholders are surgeons, a CFO, treatment coordinators, and a chief sales officer, and none of them write tickets. Requirements live in their spreadsheets, their meeting transcripts, and their workarounds. The discovery work is the work.
The dashboard is the cleanest example. Leadership ran the company on a hand-built Excel sheet with patient-to-lead links that silently broke on duplicate names. I rebuilt it where ops already lives (Google Sheets), encoded the ops manager's exact business rules as reviewable config, iterated on it live in meetings with the CSO, and backed every number with an independent verification harness. Fixing the linkage surfaced roughly $460K of patient value the old reports were quietly dropping.
The other half of forward-deployed work is enablement. One engineer doesn't scale; an organization that uses AI well does. At Hybridge I started a weekly AI Power Hour plus office hours for seven non-technical departments, built from a needs survey rather than a hype deck, with privacy and compliance addressed in every single session.
And sometimes the most valuable deliverable is the honest no. The CRM blueprint I wrote tells leadership plainly: if you won't fund ongoing engineering, the correct decision is to stay on the current vendor. Knowing when not to build is part of the job.
