The Author Is Right About Vertical AI Win, But Misses the Real Prize: Healthcare

I stumbled on this piece from Richard de Silva at Crunchbase News about the death of horizontal SaaS and the rise of vertical AI-native software. His core argument—that generic per-seat models are dying, and the next winners will be vertical specialists with proprietary data and domain expertise—is solid. I've seen this pattern play out across dozens of industries. But here's what bothered me: he frames the opportunity almost entirely around white-collar services—legal, insurance, accounting. That's the shiny object everyone's chasing. The real untapped goldmine is in frontline clinical workflows where problems are severe, fragmented, and crying out for AI-native solutions.

De Silva says AI-native software targets a $2 trillion white-collar services market and references McKinsey's $6 trillion productivity opportunity (though our data notes that figure is unverified and differs from McKinsey's known estimates). Those numbers grab headlines, but they miss the point. The most acute pain isn't in a law firm drafting contracts faster. It's in a hospital where a nurse spends 90 minutes per shift on documentation instead of patient care, or where a bedridden patient can't get a blood draw at home because no one solved the logistics ('HomeDraw'). These aren't hypothetical. On PainSignal, we track 441 healthcare-specific problems, all rated severity 5/5—the highest level of pain. When severity is that high, willingness to pay follows.

De Silva argues that defensible moats come from three D's: distribution, domain expertise, and proprietary data. I'd add a fourth: pain-first problem selection. The most durable vertical AI companies don't start with a technology or a pricing model. They start with a specific, awful problem that forces people to seek a solution. In healthcare, problems like credential verification delays ('CredentialClear') and glucometer data mismatches are so broken that even a basic AI integration can create enormous value. The data moat comes naturally when you're the first to digitize an analog process.

Where I push back on de Silva is his dismissal of horizontal categories. He writes that 'form builders, project management platforms, SMB-focused CRMs, off-the-shelf social schedulers… are compressing fast and may not recover.' That might be true for generic versions, but our data shows that horizontal tools deeply embedded in vertical workflows retain their moat. For example, 'SafeStaff ER'—a scheduling tool for emergency room nurses—scores 57/100 on our opportunity scale with severity 5/5. It's a scheduling platform, horizontal in function, but vertical in context. The switching costs aren't in the features; they're in the integration with hospital shift rules, credentialing, and compliance. De Silva's blanket claim ignores that nuance.

Another point: de Silva treats per-seat pricing as if AI killed it. He says 'the model evaporates the moment AI agents generate most of the usage.' True for generic SaaS, but usage-based and outcome-based models have existed in vertical SaaS for years. Compliance platforms, credentialing software, and revenue cycle management tools have long charged per transaction or per successful claim. AI accelerates the shift but didn't invent it. Builders in healthcare should consider hybrid models: a base subscription for the software overlay plus a per-use fee for agentic actions. That's what we see in the most promising opportunities.

De Silva correctly highlights human-in-the-loop (HITL) as a defining feature for high-stakes verticals. Healthcare is the ultimate HITL domain. A medication error from an AI agent isn't a lost contract—it's a lost life. Our data on 'DoseGuard Infusion' (severity 5/5, opportunity 62/100) shows that nurses verify every dose manually today. An AI that handles the routine checks but escalates exceptions to a human isn't just a feature; it's a requirement. Startups that design for this from day one build trust that generic AI toolkits can't replicate.

The article ends with a vision of companies that 'collapse the boundary between software and services entirely.' That's already happening in healthcare. Look at 'CodeSync' (opportunity 62/100)—a clinical coding platform that blends AI with human coders to handle the nuance of medical billing. The service layer isn't an afterthought; it's the moat. Every chart reviewed refines the model and deepens the customer relationship.

So what does this mean for indie hackers and seed investors? Don't chase the $2 trillion headline. Pick one narrow, high-severity problem in a vertical like healthcare—credentialing, infusion safety, nursing schedules—and build an AI-native solution that combines proprietary data with a human-in-the-loop design. The market is bigger than white-collar services, and the pain is real. De Silva is right about the direction, but the destination is even more valuable than he imagines.