Most people think healthcare's big AI moment is diagnosis or drug discovery. Dr. Darshak Sanghavi argues it's already happening somewhere far less glamorous — medical billing.
Darshak is Chief Medical Officer of Machinify, the AI payments platform that works with 18 of the 20 largest health plans, covers more than 170 million lives, and delivers over $4 billion in annual savings. He's also a practicing pediatric cardiologist, the first Director of Prevention and Population Health at CMMI, an early ARPA-H program manager, a former Brookings Fellow, and a bestselling author whose work has run in the New York Times and Washington Post.
In this episode, we get into the "AI war" quietly playing out between providers and payers — one side's software maximizing what gets billed, the other's reverse-engineering whether it should have been. Darshak explains where the 15% of every healthcare dollar actually goes, why "denial" means something very different inside payment integrity than it does on the news, and what he'd tell any founder trying to build into this market.
What we cover
Why the real AI business case in healthcare is back-office, not bedside
How a 500-page hospital chart used to take months to review — and what an LLM does with it now
Where Machinify's $4B in savings actually comes from (subrogation, DRGs, outpatient)
The race to "move leftward" from pay-and-chase to real-time prepayment
Why data liquidity, not technology, is the real bottleneck
What he built at CMMI: securitizing future heart attacks into a working futures market
The methodological mistakes that kill health tech pilots in front of sophisticated buyers
Why he thinks payers have a branding problem, not a value problem
Chapters
00:00 — Intro
01:08 — The provocative STAT piece: AI's real deployment is administrative
03:37 — How value-based care accidentally built an administrative empire
05:28 — What actually happens to a claim, start to finish07:56 — The downstream cost: premiums, deductibles, and his own pediatrician bill
09:47 — For founders: cheaper coders vs. solving the root problem
11:27 — Will payers pay for it? Rethinking contingency pricing
12:56 — Why he joined Machinify
15:04 — From a famous Boston trial to a bestselling book
17:41 — AI slop, and his three-step framework for writing that lands
19:45 — Building one platform out of four acquisitions
23:02 — Old world vs. new world: reviewing a sepsis claim
24:51 — Human in the loop: augment or replace?
26:20 — Good private equity vs. the Steward Healthcare playbook
27:47 — Where the $4B in savings actually comes from30:07 — The race to prepay: "moving leftward"
31:32 — Is payment integrity just a euphemism for denials?33:15 — Why data flow, not tech, is the rate-limiting step
35:22 — TEFCA, HIEs, and the information superhighway problem
37:51 — CMMI, Million Hearts, and a futures market for heart attacks
40:04 — Dallas: how data transparency fixed maternal outcomes
41:32 — The pilot questions every founder gets wrong
43:13 — Medicaid pressure, federal uncertainty, and capital paralysis
44:56 — The one thing he'd change about how CMS pays for innovation
45:35 — When value-based care actually arrives (hint: state by state)
46:52 — How payers diligence AI startups
49:02 — Optum Labs, opioid prescribing, and what models can do now
50:40 — Rebuilding trust in payers
53:45 — For clinicians who want to build or invest
54:41 — Rapid fire
Darshak's LinkedIn: https://www.linkedin.com/in/darshak-sanghavi-0427a4168/
Darshak on X: https://x.com/darshaksanghaviMachinify Website: https://www.machinify.com/
Machinify on LinkedIn: https://www.linkedin.com/company/machinify/Vik's LinkedIn: https://www.linkedin.com/in/sathvik-bilakanti-8b5344169/?skipRedirect=true
Care Shift Website: https://careshift.vc/
Vaelion Ventures Website: https://www.vaelion.vc/
