Visit lobstercap.com for more
Thanks for reading! Subscribe for free to receive new posts and support my work.
Most investors looking at insurance default to the same idea: build tools for brokers and hope the industry adopts them.
Harper took the opposite approach.
They didn’t try to automate around the industry. Instead, they’re rebuilding it from the inside.
For decades service based businesses have been off the menu for Venture Capital… but now with AI, owning the customer, the workflow, and the data end-to-end is looking really attractive.
Lobster Capital invested in Harper coming out of YC W25, and six months after signing their first customer they hit $1.1M ARR. They’re now adding over 100 customers per week with retention exceeding 98%, and the unit economics work today while getting dramatically better as automation scales.
This is the “lean AI” thesis applied to an $85B market that’s never seen real operational innovation, and it’s working faster than we expected.
Here’s why we invested…
Pivoting Away From Normal
Dakota and Tushar didn’t start with the plan to become an insurance brokerage, because no one starts with that plan.
They started where most technical founders would start: building AI tools for insurance agencies, thinking they could sell software to a fragmented market of 500,000+ brokerages.
That turned out to be a terrible idea… Insurance brokerages don’t buy software. Most run on thin margins with long sales cycles, and the 30-person shop in Toledo isn’t buying your AI copilot no matter how good the demo looks.
So they pivoted to BPO, thinking maybe the play was “we’ll do the sales work for you using AI agents.” They signed 10 agencies, learned the workflows intimately, and hit a different wall. The margins were razor thin because brokerages wouldn’t give up enough commission.
The breakthrough came when they asked themselves why they were selling insurance on behalf of these agencies when they could just get licensed themselves.
Why E&S Insurance, and Why Now
To understand why Harper is working, you need to understand what excess and surplus insurance is and why it’s been stuck in the dark ages despite being an $85B+ market.
Standard insurance covers risks that fit established underwriting criteria e.g. your typical restaurant, your average office building, a straightforward trucking operation.
E&S insurance covers everything else… the risks that are too complex, too specialized, or too unusual for standard carriers to write. e.g. trucking fleet carrying hazardous materials across state lines.
What used to be niche risks are now everyday realities. Climate change and new technologies are pushing businesses out of standard coverage and into the specialty market.
However, the market has been crippled by manual, paper-intensive workflows.
This model has persisted because the necessary technology to automate these judgment-heavy, unstructured workflows at scale did not previously exist.
Until recently. Foundation models crossed a capability threshold roughly 18 months ago that makes reliable workflow automation possible in ways it wasn’t before.
Even though everyone has access to the same foundation models, the opportunity comes from market timing, not just technical capability. Three conditions have aligned at once:
- A collapsing talent pipeline
The insurance workforce is aging out, with almost no young entrants. The average agent is in their late 50s, creating a massive labor gap.
- Carriers desperate for scalable distribution
Insurers urgently need partners who can aggregate volume and grow efficiently, and they’re actively seeking brokers who can scale.
- A broken, outdated software stack
Two legacy systems (Applied Systems and Vertafore) control 90% of the market. They’re 40+ years old, poorly integrated, and force endless manual data re-entry.
This convergence of technical capability and market desperation created a window, but it still required the right team to execute.
Most technical founders avoid starting an insurance brokerage. It requires regulatory compliance, state-by-state licensing, undocumented industry norms, and earning credibility with underwriters who’ve spent 30 years in the business. It’s slow, unsexy, and doesn’t fit the “build → launch → scale” venture script.
Dakota and Tushar had a rare mix that made it possible.
Dakota came through Goldman, Carlyle, and Coatue before co-founding Poolit, which grew to $100M AUM. He knows how to navigate regulated markets and build trust with institutional players, and he grew up around insurance, so he understood the industry wasn’t as impenetrable as it looks.
Tushar spent nearly a decade at Goldman building production ML/AI systems for commodities and asset management - environments where errors have real financial consequences.
Most importantly, they were willing to do the unglamorous work. Getting licensed in 48 states takes months. And now at Harper, everyone, including engineers (!!) must hold an insurance license.
Lean AI Model in Service Businesses
The “lean AI” model has been proven in pure software companies like Cursor, Mercor, and Lovable: high revenue per employee, minimal headcount, software-like margins.
But applying this model to insurance brokerage was genuinely novel, because traditional brokers are stuck in high-touch, labor-intensive operations.
Harper started with a clean slate and architected for AI from day one.
The result is visible in the metrics: $1M+ ARR in six months with a fraction of typical brokerage headcount. Revenue per work hour grew from $6 in October to $71 in February, which isn’t incremental improvement, it’s a structural advantage that compounds as automation improves.
The fragmentation of the E&S market means Harper doesn’t need to beat Marsh or Aon to build a massive company.
They’re not competing for Fortune 500 enterprise accounts where relationships and white-glove service matter more than operational efficiency. They’re aggregating the middle market that big brokers systematically ignore because it doesn’t fit their operating model. A towing company in Michigan with 20 trucks doesn’t merit a dedicated relationship manager at a major brokerage, but they still need complex insurance and they still compare shops on price and service quality.
Harper serves them at a fraction of the cost with faster turnaround, and there are hundreds of thousands of these businesses.
Our Investment
We’re numbers people, so here’s what mattered.
Harper signed their first customer in October 2024. In 5 months they hit $1.1M ARR and are currently adding over 100 customers per week. Today their constraint isn’t demand but operational infrastructure to handle throughput at scale.
That’s a good problem!
Today, their customer retention is running at 98.78%, which in insurance is exceptional because switching brokers is common when businesses shop for better rates.
Customers care about three things:
- speed (how fast can I get a quote)
- price (am I getting competitive rates)
- service quality (will someone help me when I have a claim).
Harper delivers on all three because of AI, but that’s implementation detail from the customer’s perspective.
When we saw retention in the high 90s, margin expansion happening in real-time, and a customer mix that spans genuinely hard-to-place risks, we knew this wasn’t a demo that worked in a controlled environment. This is a real business with improving unit economics.
Insurance brokerage economics are straightforward:
- Commissions run 10-15% of the premium that customers pay to carriers.
- Financing referral fees add another 5%, because most commercial insurance is paid annually but most businesses prefer to finance that cost over 12 months.
- Harper’s blended take rate is roughly 20% of premium volume, which is substantially better than the 2-3% they could have extracted selling BPO services to agencies.
The retention dynamics are also better than SaaS, because unlike software where customers can cancel anytime, insurance renews annually and switching brokers is painful enough that most businesses don’t do it unless they’re unhappy. If you deliver good service, customers stay.
The retention Harper is seeing creates natural compounding where revenue grows without equivalent growth in customer acquisition costs.
More customers generate more workflow data on what works (which underwriters respond quickly, which carriers give best terms for specific risks, which application formats get fastest turnaround). Better data enables better model fine-tuning, because Harper can train on proprietary outcomes rather than generic insurance knowledge.
Better models drive more automation, because tasks that used to require human judgment become predictable enough for AI. More automation increases throughput capacity, which lets them serve more customers without proportional headcount growth.
The network effects compound quietly in ways that are hard to see from the outside but create real defensibility.
The Takeaway
The broader lesson for other founders looking at AI opportunities: the winning move probably isn’t building tools for incumbents in fragmented industries.
Incumbents in fragmented markets don’t adopt software well, margins on selling to them are thin, and you’re dependent on their willingness to change.
YC’s Jared Friedman recently wrote about wanting to fund more “full-stack AI companies,” and he used a perfect example to explain the concept. Suppose you believe LLMs can automate legal work. You could build an AI agent and sell it to law firms, which is what most people do. Or you could start your own law firm, staff it with AI agents, and compete with existing law firms.
That’s going full-stack and involves targeting industries with bloated workforces, especially one dominated by slow-moving incumbents. Instead of selling to the dinosaurs, you could make them extinct.
Harper is exactly this thesis in practice.
Thanks for reading! Subscribe for free to receive new posts and support my work.
Visit lobstercap.com for more
Today’s newsletter is brought to you by Stable (YC W20).
Most founders don’t think twice about the address they list when they incorporate — until they realize it’s their home address showing up on public records.
Stable fixes that.
It gives you a real business address you can use everywhere: incorporation, IRS, banks, vendors.
It keeps your personal information private and helps your business look professional from day one.
Every piece of mail that comes to that address, Stable digitizes — with AI summaries that tell you exactly what each item is and what needs your attention, all from one secure dashboard.
Want to never think about address or mail management again? Get an address here
P.S.
Get 50% off a Grow or Scale plan for 3 months with code:LOBSTER50.