Forward Deployed Operator, Customer Operations
Confirmed on the employer's own hiring board on Nov 13, 2026. First seen by Alion on Nov 12, 2026. Harper scores A on the Alion truth index.
At a glance: San Francisco, on site, five days · $150K to $180K, or $185K to $220K at the senior level, plus 10 to 15% bonus and equity · No insurance background needed · We pay for your license
Harper is an AI-native commercial insurance brokerage, backed by Y Combinator, with offices in San Francisco and Atlanta. We do the insurance work end to end for about 1,000 new small businesses a month, with agents and people working the same accounts.
Role in one line
Own one piece of our customer operation, from doing it by hand to building the agents that run it, and keep it running.
Every day more than a hundred small businesses go from “sold” to “covered.” The gap between those words is where customers get lost. A financed deal gets marked as paid in full. A PO box gets listed as the business address. A carrier needs a document nobody asked for. You find those problems before the customer does, and then you make sure nobody has to find them by hand again.
Why now
Engineering is shipping the tools: a checklist inside every order, pay type derived instead of typed, the pending-cancellation list pulled daily, documents pre-extracted. A tool only counts when the floor uses it and the number moves. An hourly payments bot we shipped this fall went unused, because nobody owned it. Every workflow our agents take over is work our team does not have to hire for. You own one of those workflows and the number that proves it.
What you own
One workflow, end to end. The first two seats take activation checks, and payments with pending cancellations. Later: classification accuracy, certificates, intake. One workflow, one number, your name on it.
The work itself, first. Call the customer, chase the carrier, fix the record. You learn the edge cases by living them.
The agent that runs it next. Agents, internal tools and queries, built with Claude, Cursor and Harper’s own tools. Prototype in days, ship in weeks.
The line between agent and human. Some steps must stay with a licensed person. You decide which steps an agent can own, which need a licensed checkpoint, and how the system fails safely.
Adoption, not just the build. A bot nobody uses is a demo. You put it on the floor, train the team and track whether it is used.
Fewer steps. Before you automate a step, you ask whether it should exist at all.
What done means
Your workflow’s weekly number moved, and you can show the before and after.
The team relies on what you built without you in the room.
The workflow needs fewer human minutes per deal than when you started.
Your first 90 days
Day 30: you have worked the queue, your license is done or close, and the first version of a fix is in use.
Day 60: you own the workflow’s weekly number and report it every week.
Day 90: the number moved, the team relies on what you built, and you are scoping the next workflow.
Who thrives here
You have run a real operation: a queue, a book, a city, a desk. You know what a 7 AM fire feels like.
You have built something people used and can tell us the problem, the approach, the tools and the measured result.
You spend more time in Cursor or Claude Code than in spreadsheets, and you know where these tools fail.
A problem that shows up once in a hundred cases bothers you until you understand it.
You are comfortable in SQL. Python or TypeScript is a plus.
What we teach, what you bring
We teach the insurance and pay for your license. You bring operating instinct, building speed, and the ability to tell a frustrated small business owner the truth and keep their trust. Backgrounds that fit: deployment or strategic projects roles at AI companies, marketplace and fintech operations, high-volume customer operations, GTM engineering, early startup operations. Former founders welcome.
Who should not apply
You want to build product without operating. This seat runs the work too.
You want close management or a fixed spec.
You measure yourself by what shipped rather than what changed.
Pay, schedule and place
$150K to $180K, or $185K to $220K at the senior level, plus a 10 to 15% bonus tied to the workflow you own, plus equity.
In the San Francisco office Monday to Friday, starting around 7:30 to 8 AM, with regular time on our Atlanta floor. A real share of the week is calls, chasing and cleanup; the building is earned by doing the work first.
Benefits
Health, dental and vision. Breakfast, lunch and dinner at the office, commuter benefits and a gym membership. Your license is on us.
How we hire (usually inside two weeks)
A 15-minute call with a founder. We reply within 48 hours at every stage.
A build sample: 90 minutes with any AI tools you like, on a synthetic set of orders with problems hidden in it. Triage it and build one reusable check.
A Super Day on site, working practice cases beside the team. We decide within 48 hours.
How to apply
Send your resume and a short note or two-minute video on one thing you automated: the problem, what you built, and the numbers before and after. Or build to apply: pick a problem in small-business insurance operations, build something with AI tools, and send us the link. Strong builds skip straight to the Super Day.
Harper is an equal opportunity employer. If you need an accommodation at any point in the process, tell us and we will make it work. We will consider qualified applicants with arrest and conviction records in a manner consistent with the San Francisco Fair Chance Ordinance.