Same Problem, Different Waiting Room

The regulatory complexity that's supposed to keep startups out of veterinary, dental, and legal markets? It's not the barrier to entry everyone thinks it is. It's the barrier to exit.

Practitioners in all three industries are locked into software they hate. Not because the software works, but because the compliance workflows, patient records, and billing systems are so tangled together that leaving risks breaking the practice. Every manual workaround built over the past decade is now a load-bearing part of the operation.

I first noticed this as a teenager. My first job was in the local pharmacy, logging prescriptions, processing orders, managing stock. The logistics worked. The software didn't. Clunky, slow, built in the mid-nineties, full of manual steps nobody had ever fixed. The pharmacists didn't complain. They'd learned to work around it.

Years later, I visited Lupa's London HQ and heard the founders describe the exact same dynamic: qualified professionals spending a third of their time on admin, locked into legacy systems, navigating compliance workflows that hadn't been redesigned in decades. Different industry. Identical structure.

That became my Lupa case study. But the deeper I went, the more I realised this was bigger than any single industry.

The Pattern That Matters

Most regulated industries share surface-level similarities. Compliance requirements. Legacy software. Fragmented markets. That's not enough to build a thesis around. I look for a specific combination; all four conditions present simultaneously.

The practitioner is the buyer, the user, and the person whose workflow gets disrupted. There's no CTO between the software and the work. The person who has to adopt the tool is the same person whose day gets worse during the transition. Veterinary, dental, and legal all have this. Over 10,000 law firms in England and Wales, the vast majority with fewer than ten solicitors. Thousands of independent dental practices. Roughly 80% of vet practices independently owned before the corporatisation wave. Cottage industries at enormous aggregate scale.

The regulatory environment requires industry-specific data handling that horizontal tools can't replicate. Not "regulation" in the abstract. Specific, testable requirements that break general-purpose software. Controlled drug registers and RCVS audit trails in veterinary. NHS claiming integration and CQC compliance in dental. SRA-mandated matter handling and client-money accounting in legal. A general-purpose CRM doesn't fail because it's bad software. It fails because it can't pass an audit.

The existing software is genuinely irreplaceable in practice. Not because it's good, but because a decade of clinical records, billing history, and compliance documentation sits inside it, and the data isn't portable. I watched my pharmacist navigate his system through a sequence of clicks memorised by muscle memory over years. He couldn't explain why. He just knew that if he deviated, the order wouldn't process.

In verticals with genuine demand backlogs, admin time structurally caps revenue. This is clearest in veterinary, where most practices have appointment waitlists stretching weeks. Shift the clinical-to-administrative ratio from 60/40 to 80/20 and you've increased effective capacity by a third without adding headcount. That's not a software upgrade. That's a step change in practice economics.

These conditions exist beyond vet, dental, and legal. They're where I've done the deepest work, and where the pattern is most legible right now. The thesis is the pattern, not the industries. The industries are where I'm testing it.

I know why pharmacy, despite my firsthand experience, doesn't make the list: it has the compliance burden but centralised purchasing through chains and NHS frameworks. The practitioner-as-buyer dynamic is weaker. The switching costs are institutional, not personal. Accounting is similar; lock-in exists, but it's less real-time mission-critical. When one of these four conditions is missing, a company can still win. But the compounding dynamics weaken.

The Incumbents Are Comfortable

The regulatory moat hasn't just kept new entrants out. It's kept incumbents lazy.

When your customers can't easily leave and no credible competitor threatens your position, why rebuild your architecture? Why do anything other than collect maintenance revenue?

Practitioners don't hate software because they're technophobes. They hate it because they've been lied to by software vendors for twenty years. Every "seamless integration" that wasn't. Every "intuitive interface" built by someone who'd never spent a day in a clinic.

That's scar tissue, not ignorance. And it determines how the next generation of products has to sell.

Where the Economics Are Honest

The time-waste argument isn't equally clean across all three verticals.

In veterinary, freed capacity translates almost directly to more consultations. Demand is unconstrained. Every hour reclaimed from admin generates revenue.

Dental is messier. NHS contract caps on UDAs mean efficiency gains only convert to revenue on private work. The private market is growing, but for NHS-heavy practices, the unit economics argument is weaker than I'd like.

Legal is the most complex. The billable hour creates a genuine tension; some firms are incentivised to spend more time, not less. Making a lawyer 30% more efficient could look like a revenue problem under traditional partnership economics.

But the billable hour is dying. Alternative fee arrangements are growing across every practice area. The firms adopting AI aren't cutting hours; they're taking on more matters with the same headcount. The efficiency gain doesn't shrink the practice. It scales it.

The thesis is strongest in veterinary. It's real but conditional in dental and legal. That's not a weakness; it's what makes it an actual thesis instead of a slogan.

Why AI, Not Just Better Software

Here's the question that separates this from every vertical SaaS pitch deck circa 2012: could you swap "AI" for "cloud software" and make the same argument? For most of this piece, uncomfortably, yes.

What changed is two things. First, the ability to convert unstructured clinical and professional workflow (voice notes, handwritten records, free-text consultations) into structured, compliance-ready data at near-zero marginal inference cost. Rules-based software couldn't reliably do this at scale. Templates couldn't. Classic analytics couldn't at all. AI can. That's why the vertical opportunity exists now and didn't five years ago.

Second: the data flywheel. Every consultation transcribed, every claim processed, every case managed generates proprietary training data that doesn't exist on the public internet. Once established at scale, a vertical AI platform learns things about clinical workflow patterns, documentation conventions, and practice-specific operational rhythms that make it genuinely irreplaceable. Not because the AI is smarter. Because the data is proprietary and compounding.

A horizontal AI assistant can draft a generic letter. It cannot draft a client care letter that passes SRA compliance, references the correct matter details, and follows practice-area-specific conventions. A general-purpose transcription tool can capture what a vet says. It cannot structure that into a clinical record that meets RCVS standards, flag species-specific drug interactions, and auto-populate the practice management system.

That gap appears to be widening with every day of workflow data these vertical platforms collect.

What Actually Wins

Four things. In order of importance.

The data flywheel comes first. Without proprietary workflow data accumulating over time, you're a wrapper. With it, you're irreplaceable.

Replace the system of record. Build a plugin for the legacy PMS and you've inherited all its problems. The companies that compound say: we are your new operating system.

Founding teams need domain and technical depth. Pure technologists underestimate regulatory failure modes: what happens when an audit trail has a gap, what breaks when a retrospective edit isn't logged, what inspection panic looks like at 7am on a Tuesday. Pure practitioners underestimate the models. The best teams have both.

Go-to-market has to respect adoption patterns. No top-down. No PLG from a landing page. You show up at the practice, watch the workflow, earn trust from practitioners who've been burned before. These companies grow slower initially and compound faster.

Where This Breaks

Consolidation is the biggest risk. IVC Evidensia owns 1,500+ veterinary practices across Europe. Similar rollups are accelerating in dental. If corporate groups build rather than buy, the fragmented-market thesis gets harder. The counter: consolidators are the first buyers willing to pay for real technology at scale. The workflow pain is identical whether the practice is independent or part of a 500-clinic network.

The second risk: foundation models get good enough that vertical tools lose their edge. If a general-purpose model can reliably pass an RCVS audit or generate SRA-compliant documentation out of the box, the vertical premium erodes. I don't think that's where this is heading. But I hold the thesis with open hands.

The Lens

This is how I evaluate every software opportunity in a regulated industry now. Not "is this market big?", because every market is big if you squint. Not "is this software bad?", because most software is bad.

Three questions. Is the regulatory complexity functioning as an exit barrier rather than an entry barrier? Are practitioners locked in by load-bearing workarounds rather than genuine product loyalty? Does AI unlock a data flywheel that makes the product more defensible with every hour of use?

Where all three hold, the opportunity is structural.

Boring industries. Brittle incumbents. Exit-barrier lock-in. And a very specific kind of company, quiet, early, compounding in markets most investors scroll past, building the replacement while nobody's watching.