A 6-person team shipped healthcare prior-auth AI agents in 3 months — here's the playbook
alex_verem · x · 2026-10-15
A PE operating partner asked the author's team to put AI agents inside a portfolio company's prior-authorization platform, where patient data flows to insurers under HIPAA. The old RPA bots broke so often on insurer portals that coordinators went back to manual submission.
The playbook they share:
- Instead of hardcoding one path and patching it forever, or rebuilding the whole platform, they picked the most painful piece only: portal automation. Case management, EHR links and fax stayed on the legacy system.
- Month one wasn't about building a product — it was proving one thing on one live insurer portal: an agent could operate it from inside the coordinator's own logged-in browser session.
Three months later, a team of six shipped a production browser extension running those submissions inside coordinators' browsers. Key lesson: prove one thing on one real environment first, then expand.
More from coding & agent
- AI coding agents on production codebases: 4 setups tried, 3 failed, one workflow stuck — paulabartabajo_ · 2026-10-15
- Sarvam launches Voice Agents MCP to build calling agents from your coding tool — itsOmSarraf_ · 2026-10-15
- XERJ indexes code, docs and PDFs in one command, saving agents 40x tokens vs grep — xerj-org · 2026-10-15
- OpenAI caught a rogue agent in 15 minutes — but it ran 2.5 more hours until a human stopped it — xaviersilient67 · 2026-10-15
- Don't blame the agent: match the environment to the task, then instrument outcomes and infra failures — bibryam · 2026-10-15
- LLM agents autonomously design a SAR ADC on SKY130 from schematic to signoff, paper finds — IgorCarron · 2026-10-15