Healthcare needs AI that works for the overlooked 87%

Three out of every four small clinics we track cite 'cost of new technology' as their top barrier to adopting AI. That's according to PainSignal's database of over 300 healthcare administration problems, where small practices (fewer than 10 staff) consistently report training time as a severity 3.8 out of 5 issue. So when Lindsay Stanley at CB Insights profiles Penguin AI — a no-code agentic workflow builder — I can't help but think the interview tells only half the story.

The interview is a classic AI100 finalist feature: why your company matters, what sets you apart. Glenn Herzberg, Penguin AI's head of marketing, makes the case that healthcare organizations are drowning in manual work — prior authorizations, care coordination, clinical documentation — and that most AI tools are either black boxes or too burdensome. Penguin AI's solution, Gwen, lets frontline workers build and deploy workflows without writing code. It sounds great. But our data suggests two massive gaps in this narrative.

First, the scale of the problem is real but often overstated. The article claims administrative tasks eat up 'trillions of dollars.' Total U.S. healthcare spending is about $4.5 trillion, and administrative costs are estimated at around $265 billion annually (HHS 2019). Still enormous, but not trillions. Our own tracking shows prior authorization delays alone cost roughly $30 billion annually in overhead — a massive inefficiency that any automation should target. But hyperbole doesn't help anyone.

Second, and more critically, the interview glosses over interoperability. We track 203 problems tagged 'data interoperability' across healthcare, with EHR data silos scoring a severity of 4.5 out of 5 — the highest we've recorded in any category. Even a beautifully designed no-code builder like Gwen is useless if the underlying data is trapped in incompatible systems. Smaller clinics, especially those with limited IT budgets, face this wall daily. They can't just plug in a workflow builder; they first need to untangle decades of fragmented data storage.

And that brings me back to the small clinics. We track 87 distinct problems from providers with fewer than 10 staff. Their top complaint isn't 'we lack automation tools' — it's 'we can't afford the change.' New tech means training time, integration headaches, and ongoing maintenance. Even a no-code solution requires someone to learn it, champion it, and troubleshoot it. For a three-doctor practice with one overworked admin, that's a non-starter.

None of this means Penguin AI's approach is wrong. In fact, the concept of agentic workflows built by the people who understand the work is compelling. But the real innovation in healthcare AI won't come from just making the builder easy — it'll come from solving the data plumbing and making adoption painless for the 87% of small providers who are currently priced out of the conversation.

So if you're building in this space — and I know many of you are — look at what the big players miss. The data is clear: interop severity 4.5, small clinic cost barrier 4.1, training burden 3.8. Those are the numbers that matter. Build for them.

Featured data from PainSignal's healthcare dataset.