Vertical AI in Legal & Dental
Everyone talks about vertical AI, but most people get the thesis wrong.
The standard pitch is "AI for X industry." But that's not a thesis – it's a category. The actual thesis is about why certain verticals are defensible and others aren't. It comes down to three things: regulatory complexity as a moat, domain expertise as a barrier, and workflow integration as lock-in.
I spent December analysing two Seedcamp portfolio companies that embody this framework. Both are attacking regulated professional services. Both have teams with deep domain pedigree. Both are building workflow-specific tools rather than horizontal features.
The question isn't whether vertical AI works. It's whether these specific teams, in these specific markets, at this specific moment, are the right bet.
Wexler: AI for Litigation
Wexler – Stage: Seed | HQ: London | Sector: AI / LegalTech
Building AI infrastructure for complex litigation – chronologies, fact extraction, and real-time testimony verification for elite law firms.
Complex litigation is where law meets data science at massive scale. A single commercial dispute can generate hundreds of thousands of documents. Establishing basic facts – who said what, when, and whether it contradicts prior testimony – still relies on junior lawyers billing thousands of hours for manual review.
Wexler replaces this with AI that operates at fact level, not document level. Most legal AI tools help you find documents. Wexler extracts, analyses, and cross-references the actual information inside them. It thinks like a litigator but operates at machine scale.
The founding story is unconventional. Gregory Mostyn isn't a lawyer – he's a repeat entrepreneur whose father is a recently retired High Court Judge, whose brother is a partner at Cleary Gottlieb, and whose stepmother is a barrister. Law runs in his blood without the professional constraints that sometimes limit domain experts from building transformative products.
"Facts are the backbone of a system that could help solve complex disputes."
The early traction validates the approach. Clifford Chance has embedded Wexler across its dispute resolution practice. HSF Kramer, Goodwin Procter, Burges Salmon, and Addleshaw Goddard are customers. These are among the world's most demanding law firms – their adoption signals both product quality and enterprise readiness.
The metrics tell the story: 20x ARR growth since pre-seed, 75% reduction in manual review time. Chronologies that once took days now take about an hour. They've had profitable months at seed stage.
The Bear Case
eDiscovery incumbents like Relativity and Reveal are adding AI capabilities. Horizontal legal AI players may move into litigation. The market will attract well-funded competition.
But Wexler's vertical depth creates meaningful differentiation. The platform operates at fact level rather than document level – that's not a feature competitors can bolt on. It requires rebuilding from the architecture up.
What I'd Want to Know
Three questions I couldn't answer from public sources. First: how dependent is current ARR on Clifford Chance? Elite firm customers provide credibility but can create concentration risk.
Second: the open question is whether pricing power compounds with workflow depth or caps out at matter-level ROI. Per-seat scales with firm size. Per-matter scales with case volume. The model they choose will define margin trajectory.
Third: what's the false positive/negative rate on real-time fact-checking? In litigation, errors have consequences. This is the trust barrier that determines whether Wexler becomes essential infrastructure or a tool that still gets checked.
Verdict: Bullish. This is a conviction bet on vertical AI infrastructure that could change the economics of litigation at the world's most important law firms. The unit economics and customer quality suggest exceptional product-market fit.
La Fraise: AI for Dental
La Fraise Pro – Stage: Seed | HQ: Paris | Sector: AI / Healthcare
AI agents that transform dental practice management – automating administrative tasks, patient communication, and treatment plan follow-up.
European dental care is a €25B+ market characterised by high fragmentation and minimal technology adoption. Most practices run on legacy software or manual processes. France alone has 45,000+ practicing dentists and over 1,300 dental centres. The market reached €16B in 2024, growing 5.3% year-on-year. Yet tech adoption remains remarkably low.
The problem La Fraise attacks is specific and measurable: 50% of recommended dental treatment plans in France are never completed. This isn't primarily a clinical issue – it's an administrative and communication failure. Patients don't understand their options, can't navigate financing, or simply fall through the cracks of manual follow-up processes.
La Fraise reduces this information asymmetry and administrative friction. Early deployments show practitioners gaining 60+ minutes daily and 20% improvement in treatment-plan acceptance. That's quantifiable ROI that should accelerate sales cycles and reduce churn.
The founding team's pedigree is exceptional. Four co-founders spent 8+ years building Doctolib's product, sales, and operations – one of Europe's largest verticalised software businesses. The fifth, Arnaud Assous, is a serial entrepreneur who built and sold Camshop France and has investment banking experience at Lincoln International and BNP Paribas.
Critically, they've chosen to start in dental – a vertical Doctolib itself largely bypassed. They're not competing with their former employer; they're applying proven playbooks to an adjacent, underserved market.
The Bear Case
Dental practitioners are notoriously conservative buyers. AI adoption in healthcare remains nascent, and the regulatory environment continues to evolve. DentalMonitoring, Dental Pilote, and US players like Pearl and Overjet are all building in adjacent spaces. Doctolib itself could expand into dental.
But La Fraise's focus on administrative AI (versus clinical) reduces regulatory burden. And the team's Doctolib experience provides a proven playbook for winning trust in exactly this type of buyer.
What I'd Want to Know
Three questions worth digging into. First: the unit economics question. With 1,200 dentists on the platform, does LTV/CAC improve as product depth increases, or does the SMB nature of dental practices cap expansion revenue? Churn dynamics will tell the story.
Second: what's the real relationship with Doctolib? Are they genuinely agnostic, or is there a partnership or acquisition angle? The founders' relationships could be a distribution unlock or a competitive risk.
Third: what's the actual expansion playbook? Germany, Spain, Italy, and UK are mentioned as targets, but dental regulations vary significantly across Europe. Country-specific adaptations matter.
Verdict: Bullish. This is a conviction bet on exceptional operators attacking a regulated, overlooked sector with AI-native infrastructure. The opportunity to back founders who've already built at European scale is rare. The €3.2M round alongside 20VC and Kima Ventures signals broad conviction.
The Pattern
Both companies share structural advantages worth noting.
Healthcare and legal are heavily regulated. That's a feature, not a bug. Regulation creates barriers to entry for horizontal players who can't navigate compliance complexity. Both tools embed deeply into existing processes rather than asking practitioners to change how they work – La Fraise integrates with 100% of existing dental software, Wexler plugs into how litigators already build cases.
Neither team is learning the industry from scratch. The Doctolib alumni know healthcare software sales cycles. Mostyn's family connections give him intimate access to how litigation actually works. And both can point to specific metrics – hours saved, revenue increased, review time reduced. In enterprise sales, quantifiable value accelerates cycles and reduces churn.
That's not "AI for X." That's purpose-built tools, built by domain experts, for regulated industries with high switching costs.
The question isn't whether these companies will face competition. They will. The question is whether their head start, team quality, and workflow depth create enough defensibility to win their respective markets.
I think they're positioned to.