Can AI solve the economic frictions of healthcare? In a new video, FTV Capital Partner Alex Mason digs in.
His view: AI adoption scales in healthcare when three things are true: the risk of failure is lower, the ROI is provable and the workflow is broken.
Take prior authorization (a pre-approval step before providers can perform a service or treatment) — one of healthcare’s most infamous bottlenecks, and an example that checks all three boxes.
Major insurers last year committed to simplifying prior authorization, and in just a few months’ time, starting January 2027, electronic prior authorization via APIs will become a requirement for many payers.
The conditions might seem set for technology, specifically AI, to deliver real value. But deploying AI in healthcare is challenging on many levels, and a major bottleneck remains.
Beyond prior auth, Alex digs into why the healthcare market operates exactly as it’s designed, and how AI is beginning to create a better version of a bad system.
Full Video Transcript
It is important to understand that the healthcare market operates today exactly how it’s designed. It’s just designed really, really poorly. You have to understand the friction points in healthcare to really understand both opportunity and risk. We’re at this fascinating frontier of rapid and accelerating pace of change in AI, stuck in a market that has been archaic in its practices forever.
AI adoption scales in healthcare when three things are true: the risk of failure is lower, the ROI provable, the workflow is broken.
Take prior authorization right now. Completing a prior auth request can take days. Providers spend hours building clinical documentation, navigating payer portals, following up by phone or fax. AI is attacking this problem in a few ways. The most mature approach pulls structured clinical data directly from the EHR, matches it against payer criteria, and auto-generates and submits the authorization request. This compresses a massive amount of time and inefficiency — hours of staff work — into minutes.
What AI cannot yet do is accelerate the payer’s response time. This remains the biggest bottleneck. The submission is faster, but the clock still runs on the payer’s end. So AI cannot bridge the economic structural divide that exists. But AI can create a better version of a bad system.
