The archive · Money & Fintech · Product decision · 2025–2026
Solva bets insurers will trust verifiable AI agents; $6M seed, $245K ARR in 10 weeks
Ex-InsurTech founders build 'Harvey for insurance': AI agents that automate claims and stop wrongful payouts, backed by Paul Graham.
Solva
What the business is
AI claims-automation software for insurers: agents that sit on top of existing claims systems, detect incomplete evidence, flag policy breaches, and stop wrongful payouts.
Starting capital：$6M seed (First Round, YC, SV Angel, angels incl. Paul Graham)
How it started
Herman Båverud Olsson (ex-Lassie, Europe's fastest-growing InsurTech), Linus Malmén (co-founder of EU-licensed neobank Zaver, €1B+ in annual processing) and Sorena Amini (Citi, Fidelio) had spent a decade inside regulated finance. They saw claims handling still running on 2000-era tools and founded Solva to modernize it.
What happened
Solva went through YC's Summer 2025 batch, raised a $6M seed from First Round, YC, SV Angel and angels including Paul Graham, and launched with what it calls 'Harvey for insurance': verifiable AI agents that detect incomplete evidence, expose fraudulent behavior, and deliver source-cited claim recommendations. Within 10 weeks of launch it reported $245K in annual recurring revenue.
How it ended up
Live and scaling — a 15-person team in New York as of September 2026, pitching itself as 'the AI operating system to reinvent insurance, starting with claims.'
Background
Solva's bet: insurers pay out $6.5 trillion a year, yet claims handling still runs on 2000-era admin tools, and manual errors cause an estimated $200B in wrongful payouts plus $75B in administrative costs annually. The founders — who had scaled Europe's fastest-growing InsurTech and built an EU-licensed neobank — concluded that claims was the industry's most painful, expensive, and automatable process.
The product is deliberately non-disruptive: an AI layer that runs on top of any existing claims stack, detects incomplete evidence, identifies policy-term breaches, exposes fraud, and delivers source-cited recommendations an adjuster can verify. That verifiability, not raw automation, is the pitch — the difference from black-box AI tools in a regulated industry.
After YC's Summer 2025 batch and a $6M seed from First Round, YC, SV Angel and angels including Paul Graham, Solva reported $245K ARR within 10 weeks of launch — the figure investors cite as evidence that even conservative insurers will pay agents that directly save money. It remains early-stage, but the team's thesis is that claims is the entry point into a much larger 'AI operating system' for insurance.
What has to be true
- The wedge is the dullest, most expensive task: claims handling is where insurance loses $200B+ a year, so even a small productivity gain justifies a subscription.
- Trust is the moat in regulated industries: source-cited, verifiable recommendations let insurers test AI without surrendering control — faster than black-box tools ever could.
- Domain pedigree sells where software can't: founders who scaled an InsurTech and ran a licensed bank carry the credibility that opens carrier doors.
- Working with legacy stacks instead of replacing them lowers the buyer's switching cost, the classic wedge into a conservative market.
What can be applied
In regulated industries, trust beats speed: the product that gets in works with legacy systems and proves every recommendation, not the one promising replacement.
Aftermath
As of September 2026 Solva is active and scaling from New York with a 15-person team, selling claims automation to carriers and MGAs and positioning itself as the 'AI operating system' for insurance. It is backed by $6M from First Round, YC, SV Angel and angels including Paul Graham, and its early $245K ARR in ten weeks remains the headline traction metric.
Sources
- Launch YC: Solva — Harvey for insurance
- Solva: Automates insurance claims and stops incorrect payouts
- Two Key Trends in AI Entrepreneurship Observed from 169 Startups
- Solva Secures $6M to Rebuild Insurance Claims with AI, Backed by Paul Graham
spotted an error? The archive wants to know.
Your turn
You just read one. Describe what you are building, and see who is betting on the same thing.
Free account · 3 free questions · no card