Notes from the retrieval floor.
Engineering write-ups, field reports from our specialists, and the occasional opinion about how health data should move.
How we taught a voice agent to wait on hold without losing its place
Front desks put you on hold. A lot. Our first agents treated silence as the end of the call. Here's how we built hold detection, IVR navigation and the "call me back" fallback that cut abandoned calls by 71%.
Read the postProvenance is not a footnote: how we attach a source to every FHIR field
Why we store retrieval method and confidence alongside every value, and what it costs.
The 40-year-old paper chart in a Vermont basement
A retrieval specialist on the longest fax we've ever received, and why it mattered for a disability claim.
What TEFCA changes for patient-mediated access, and what it doesn't
A plain-English reading of the US framework, and where agents still have to fill the gaps.
Our browser agent broke on 212 portals in one night. Here's what we learned.
A vendor shipped a redesign. Our recovery took 14 hours. The post-mortem, and the self-healing selectors we built after.
Nolqa backs Retrievant with $555K to put retrieval agents in front of every care site
Why we started in Paris, why Boston came next, and what Nolqa’s investment lets us build.
Should an AI say it's an AI on a phone call? Yes. Every time.
Our disclosure policy, the research behind it, and what happened to completion rates when we adopted it.
One email a month. No more.
New posts, coverage updates, and what our agents learned this month.