Want to know if this job is worth applying to?
San Francisco, California, United States
Onsite
Harper is an AI-native commercial insurance company in San Francisco. We're not bolting AI onto insurance — we're rebuilding the entire business as software, on a simple bet: turning expert human judgment into compute is one of the largest transitions left to make, and a trillion-dollar industry still run 90% by hand is the place to prove it. We've grown ~100x in the last year and we move at that speed — on-site, in person, long days, very high standards. Almost no one joins Harper for insurance; they join to build the company that replaces how it works.
A routing rule at Harper can move thousands of accounts. If the risk judgment behind it is vague, unsupported, or wrong, scale multiplies the mistake. Great underwriting judgment exists — it's just trapped in individual inboxes, one-off decisions, messy submissions, and tribal knowledge, most of all in specialty commercial and E&S, where risk moves through PDFs, ACORDs, loss runs, supplementals, and broker notes that may or may not answer what an underwriter needs to know. You make that judgment explicit, source-backed, and safe to apply at scale.
This is Harper's in-house underwriting encyclopedia, not an underwriting desk. You do not approve or decline individual accounts. You do not own market selection, the submission narrative, negotiation, or the commercial recommendation on any account; the licensed account owner keeps those. You own the risk judgment itself: the frameworks, the routing constraints, the appetite rules, the escalation conditions, and the source evidence behind each one. You don't need to write code. You do need to be the underwriter who sees a bad submission and thinks: what should intake have asked, what should the system have extracted, what rule would have caught this, what should we never send a carrier again.
Why now: Harper already has the data, live submissions, carrier outcomes, and systems that act at scale. The missing piece is senior judgment that can explain why one risk is strong, another is weak, and a third needs different information before anyone should make a call — and can write that down so a system can apply it and a person can audit it.
Review live risk with depth. Read live accounts and system outputs to catch missing context, weak assumptions, incomplete exposure analysis, and false confidence.
Build the underwriting frameworks. Define how Harper reasons about operations, exposures, controls, loss history, coverage structure, and material risk differences — with positive examples, counterexamples, and the evidence that separates them.
Set routing constraints. Define which risk facts permit a route, block a route, or require senior review, with the source and rationale attached to each.
Encode judgment into the product. Work directly with product and engineering to turn how you think into intake questions, checklists, appetite rules, escalation conditions, quality tests, and instructions an AI can run.
Reason about carrier fit. Explain which markets fit which risk shapes, what evidence supports the fit, and where stated appetite breaks in practice.
Audit what the system decided. Test whether routing, risk summaries, and recommendations are faithful to the source record and the underwriting standard, and trace every correction back to the rule or source-reading failure that caused it.
Teach the judgment and close the loop. Coach placement specialists, account teams, operators, and product partners on how strong underwriting reasoning works, and tie quotes, declines, binds, losses, and corrections back to the framework they confirm or change.
You have 4+ years of commercial underwriting with meaningful specialty commercial or excess and surplus exposure.
You've handled messy risks and imperfect submissions at an MGA, MGU, wholesaler, specialty carrier, program administrator, or similar — not only clean admitted-market accounts.
You have breadth across commercial lines or risk categories: general liability, property, workers' comp, commercial auto, excess/umbrella, professional and E&O, contractors, habitational, restaurants, transportation, or other specialty classes.
You can explain not only your decision but the factors and thresholds that produced it, and you can separate carrier folklore from evidence.
You can express underwriting judgment as source-backed rules, examples, counterexamples, and escalation conditions that a system can apply and a person can audit.
You treat AI as leverage — you're product-curious, you'll sit with engineers and pressure-test system outputs, and you want your judgment to scale across thousands of submissions instead of the one in front of you.
You hold, or can quickly earn, the insurance licenses or credentials the work requires (we pay for it).
This job is on-site in New York or San Francisco, Monday through Friday, long days, in the building with product, engineering, intake, and the placement desk. You'll help build the system while it is already quoting live business, so there is context switching, ambiguity, and a steady stream of problems that don't fit a job description. You'll review live cases most days and you'll write more than most underwriters ever have: frameworks, rules, counterexamples, escalation conditions. Nothing is done until someone else can apply it and check it against the source.
This is not a traditional underwriting seat, and if that's what you want, don't apply. If you want authority to approve or decline individual accounts, if you want to own customer negotiation or placement strategy, if you prefer clean admitted risks and a fixed manual to ambiguous specialty work, or if you rely on unwritten instinct and don't want your reasoning made inspectable, this seat will frustrate you. Advising from a distance without getting into the submission, the system output, and the product logic doesn't work here, and neither does a remote or hybrid default. If the intensity reads as a cost you'd rather avoid, this isn't the seat. If it reads as the point, keep going.
Salary: 125,000–175,000 base, plus performance bonus and equity.
Equity: Yes.
Location: On-site, New York or San Francisco. Based there or willing to relocate.
Schedule: Monday–Friday, in-office hours that match the rest of the company. The hours are long.
Licensing: Required commercial insurance licenses vary by state; Harper pays for training and licensing.
Health, dental, and vision insurance
Meals provided — breakfast, lunch, and dinner
Snacks, drinks, and coffee stocked daily
Commuter benefits
Free gym membership
Team call — 30 minutes on Zoom. You meet the team, we meet you.
Office session — about two hours in person. You meet more of the team and work a real problem with us.
Super day — how you actually operate, in real time. We decide after this.
The process moves as fast as you do.
Send your resume and one underwriting example: a risk where your first read changed after better information arrived — which facts mattered, which didn't, and how you would turn that judgment into a repeatable Harper framework.
