TP493: Ghost Networks and the Reflex to Automate

Touch Point Podcast
Touch Point Podcast
TP493: Ghost Networks and the Reflex to Automate
Loading
/

Chris Boyer and Reed Smith bring in two people who worked the problem from the inside. Chris Hemphill of Modular Feedback, who builds AI for a living, and Heather Nairn, a healthcare economist who reads this as an access problem first.

The reflex across the industry is to point AI at the mess. Standardize the data, set some agents loose, let the model sort it out. Hemphill and Nairn tested that reflex against a plain deterministic workflow on exactly this job. The workflow won on accuracy, on speed and on cost. Their point is not that AI is useless here. It is that the most useful skill in this work is knowing when not to reach for it.

The deeper problem is structural. Provider data is a commodity. Every payer and every health system chases the same handful of fields, guards its copy as proprietary, and rebuilds the same record in parallel. Every cycle spent on that is a cycle not spent on the access work that moves outcomes. Transportation, care coordination, the patient in crisis who just needs a number that connects.

In this episode, Chris and Reed cover:

  • Why a directory of five public facts can stay wrong for 540 days at a stretch
  • What makes provider data a commodity, and why treating it as proprietary keeps it broken
  • The behavioral health networks where more than 80 percent of listings lead nowhere
  • Why a deterministic workflow beat AI on accuracy, speed and cost for this job
  • The difference between using AI to find a problem and trusting it to fix one
  • Where the cycles freed from directory cleanup could actually go

Mentions from the Show:

Sponsors