a16z maps out how AI can actually sell to healthcare

Healthcare is running out of workers. According to a16z, the average hospital turned over 100% of its workforce in the last five years, while labor costs jumped roughly 21% over the past three.

Yet, doctors and nurses hate buying new software. They already spend half their day wrestling with clinical databases, and nobody wants another tool that demands training and changes their daily workflow.

To figure out what hospitals will actually pay for, venture firm a16z mapped out an enterprise healthcare AI playbook using the classic Jobs to be Done lens.

Hiring AI workers instead of buying software

The core thesis is simple: don't sell software to employees, sell digital headcount to the enterprise.

Healthcare providers are much more receptive to "AI staff members" in domains where human labor is painfully expensive and turnover is rampant. Medical scribing is already seeing massive AI uptake precisely because it replaces manual transcription without forcing doctors to learn a complex new interface.

The second hurdle is delivering a clear 10x performance leap over status quo human processes.

Take prior authorization as an example. An AMA survey revealed that physicians and their teams spend roughly 14 hours every single week dealing with authorizations. Meanwhile, the GAO estimated that federal agencies made upwards of $128 billion in improper Medicare and Medicaid payments in 2022. That kind of administrative nightmare is primed for an AI overhaul.

The wedge is not the whole game

Selling into hospitals also requires navigating heavy regulatory boundaries.

While frameworks like the FDA SaMD and 510K pathways exist, non-clinical enterprise tools feel far safer to C-suites than diagnostic algorithms. For clinical tasks, founders still have to keep humans firmly in the loop to gain institutional trust.

More critically, a clever point solution won't survive long on its own.

a16z noted that healthcare buyers are aggressively consolidating their vendor rosters. They do not want to manage dozens of separate startups for different tasks; they want single partners capable of covering broad workflows across the enterprise.

Why it matters

Selling enterprise healthcare tech has historically been a graveyard for venture-backed startups because altering clinician behavior is brutally hard.

We think framing AI solutions as digital labor rather than workflow software is a brilliant commercial shift. It allows hospital CFOs to compare AI directly against runaway staffing costs rather than squeezing it into an already strained IT budget.

The real risk, however, comes right after product-market fit. Winning a single wedge like scribing or billing audits only gets you in the door.

Because hospital systems are actively cutting down vendor sprawl, point tools risk being commoditized or replaced unless they quickly roll out integrated suites that span multiple enterprise workflows.