Harper is an AI-native commercial insurance brokerage for small businesses. We built our own software for nearly every step, so our people spend their time where trust matters. We use AI to earn that trust.
Our AI can only broker as well as the judgment it learns from. You make expert brokerage judgment teachable, testable, and reusable across Harper.
A thin submission for a contractor with an unusual subcontracting setup gets placed badly. You trace why, write the decision standard that would have caught it, turn the case into an AI evaluation, and build a worked example for the placement team. The next ten accounts like it are handled right, by people and by the system.
At most brokerages, the best judgment lives in a few senior heads and leaves when they do. Here you turn it into the workflows, evaluations, and training the whole operation runs on.
Encode brokerage judgment. Turn complex commercial decisions into workflows, decision standards, playbooks, and escalation rules without flattening the judgment that matters.
Define product requirements. Work with product and engineering on which brokerage problems to solve and what the product must understand.
Train and evaluate AI. Build realistic cases, grading standards, and failure examples, and turn every material failure into a permanent evaluation.
Design brokerage workflows. Define how information, decisions, and handoffs should move, and test whether the workflow produces sound judgment.
Own the learning standard. Create the cases and worked examples that make expert judgment teachable, with learning and enablement delivering them.
Run the expert feedback loop. Stay close to frontline cases, separate one-offs from patterns, and keep every decision traceable to its source evidence.
You have substantial commercial, wholesale, or placement brokerage experience across complex small and middle-market accounts.
You understand GL, property, workers' comp, commercial auto, umbrella, and specialty coverage well enough to spot the material facts and exceptions.
You can turn tacit judgment into precise instructions and reviewable evaluations, and you are curious about AI and its failures.
If you meet most of this, apply.
This job does not sell insurance, own a book, carry a quota, or make final placement decisions. You bring the standard and the evidence; product owns the roadmap and account owners keep the final call. When a familiar habit and the evidence disagree, the evidence wins.
This is not the seat if you want to personally produce and bind accounts, if you would rather keep judgment in your head than in a system, or if you need a remote or hybrid schedule.
It is the seat if defining how an AI-native brokerage should broker sounds like the work you want.
Harper is an equal opportunity employer.
A 30-minute team call on Zoom.
An office session of about two hours, working a real brokerage case with us.
A super day in the office, showing how you operate in real time.
Pay: $140,000 to $210,000 base. Bonus and equity eligible.
Where: On-site, five days a week, in the room with the team. San Francisco at 425 Market Street, or New York at 12 E 49th St.
Hours: 5am to 7pm PT, Monday through Friday.
License: Commercial insurance licenses as the work requires. Harper pays for the training and the exam.
Benefits: Health, dental, and vision. San Francisco adds three meals a day in the office, commuter benefits, and a gym membership.
Send your resume and a short example of expert judgment you made reusable. Tell us the brokerage problem, why the obvious rule fell short, what you built, and how you knew it worked.