The Problem

36 million businesses in America need insurance—it's not optional. 77% are underinsured. 40% have no coverage at all. The distribution system failed them: too slow, too opaque, too confusing.

Over 90% of commercial insurance is still human-led. We're building the inverse: 90%+ AI-led, pushing toward the higher 90s. To do that, everything we do has to become legible.

Today, most of Harper's operating knowledge lives in people's heads—how a top rep prioritizes quotes, how service handles edge cases, how market routing actually works, what a customer really means when they push back at bind, why a workflow changed yesterday. That works at small scale. It breaks at ~1,000 new customers a month.

We need someone to turn tribal knowledge into operating memory—and to get the rest of the company running against it.

The Thesis

AI doesn't magically understand a company. It only works when the business is documented clearly enough for systems to retrieve the right context, recognize the workflow, handle the edge cases, and escalate when human judgment is needed.

The next bottleneck at Harper isn't engineering. It's knowledge—and the speed at which people can absorb it. Every process that lives only in someone's head is a future failure mode. Every undocumented edge case is another rework. Every AI-generated playbook that lands in chat and never gets operationalized is throughput we left on the floor.

You'll turn messy operating reality into structured, AI-legible knowledge—and make sure the people meant to act on it actually do.

The Role

Two tracks, running in parallel. You'll sit at the intersection of Operations, Engineering, and RevOps.

Train the system. You embed with sales, intake, service, placements, and renewals. You sit with operators, listen to calls, study workflows, and turn what you find into source-of-truth docs, SOPs, playbooks, decision logs, system boundary docs, glossaries, and skills the agents can call. You partner with engineering on the automation layer so docs stay alive instead of going stale.

Train the people. When a playbook gets shipped, you're the one who turns it into an executable plan—named owners, first three moves, rollout cadence. You build the onboarding paths and setup scripts that get a new hire into Cursor, Claude Code, and the harness within a week. You run cohort rollouts, drive adoption, and make activity visible.

You don't need to be an engineer. You do need to be exceptional at using AI tools—Cursor, Claude Code, MCP servers, agent memory files, Granola, structured prompting—to turn raw context into reusable work product, and reusable work product into behavior change. You'll partner directly with the CEO when extraction calls for it.

This is not a note-taking role. This is not "make the Notion pretty." This is not corporate L&D. There is no LMS, no slide deck, no e-learning project. This is an operating role for someone who can walk into ambiguity, find the hidden logic, turn it into systems—and get the rest of the company to run against those systems at AI speed.

What You'll Do

Capture tribal knowledge — Interview operators, shadow workflows, sit with teams, and document what people "just know"
Build operating memory — Turn transcripts, Slack threads, Looms, and one-off explanations into source-of-truth docs, decision logs, playbooks, and process maps
Author skills and maintain system boundary docs — Graduate stabilized rules into skills the agents can call. Hold the canonical answers for what each internal system does, where one stops and another begins, and what gets misunderstood.
Translate AI playbooks into executable plans — When a rich AI-generated playbook lands, you're the one who distills it: owners, first moves, rollout. The plan doesn't run itself.
Drive adoption — Onboarding scripts, cohort rollouts, setup paths for Cursor/Claude Code/MCP. Make activity visible so we know who's actually in the harness.
Make meetings AI-legible — Shape conversations in real time so they produce useful artifacts: decisions, owners, definitions, edge cases, unresolved questions, next steps
Find the edge cases — Document where workflows break: reworks, escalations, stale quotes, market follow-ups, payment/binder gaps, customer confusion
Maintain the knowledge base — Keep docs current, assign owners, kill stale guidance, partner with engineering on refresh automations so freshness isn't manual
Turn repeated problems into systems — If the same issue happens three times, it becomes a playbook, a QA check, a skill, or a product requirement

You Might Be a Fit If…

You're an exceptional writer and synthesizer
You're curious about how organizations actually work, and you like sitting with operators
You can turn a messy transcript into a clear operating doc—and a clear doc into something the team actually runs against
You're AI-native in practice—not just "I use ChatGPT," but you have taste for when an output is structurally wrong, not just stylistically off
You know how to prompt for extraction (decisions, contradictions, owners, edge cases), not just summarization
You understand that good AI output depends on good context, and you know how to engineer context upstream
You can take a 50-page AI playbook and produce a 5-step executable plan with named owners and a date
You're structured but not bureaucratic
You care whether documentation changes behavior, not whether it looks polished
You can operate in chaos without becoming chaotic
You're low-ego, persistent, and allergic to "someone should probably document that"

Backgrounds That Could Work

We're open-minded. Strong candidates might come from modern L&D or enablement at an AI-native company, qualitative or academic research, ethnography, instructional or curriculum design, knowledge management, product ops, technical writing, research ops, implementation, chief of staff roles, library and information science, AI ops / human-in-the-loop work, or startup operations roles where the systems were messy and someone had to make them legible.

The exact background matters less than the ability to extract knowledge, impose useful structure, use AI tools well, and ship artifacts that change how people work.

Requirements

3–8 years in L&D, enablement, research, product ops, knowledge management, technical writing, implementation, chief of staff work, qualitative research, instructional design, or startup operations
Exceptional written communication
Strong AI-tool fluency in practice: Cursor, Claude Code, MCP servers, agent memory files, Granola, structured prompting, AI-assisted synthesis
Demonstrated ability to interview stakeholders and extract operational detail
Ability to turn messy conversations into clear decisions, workflows, and source-of-truth documentation
Strong information-architecture instincts
Track record of running an adoption rollout that actually changed how a team worked
Comfort in a fast-moving, ambiguous startup
Based in San Francisco or willing to relocate

Nice to Have

Experience at an AI-native or developer-tools company
Background in ethnography, sociology, anthropology, curriculum design, or library/information science
Experience authoring Claude skills, agent skills, or similar capability modules
Experience with RAG/search systems, data labeling, evals on knowledge systems, or human-in-the-loop QA
Experience working with engineering on refresh automations on top of living docs
Experience translating operator feedback into product requirements
Experience with taxonomy, metadata, tagging, or content governance
Experience in insurance, fintech, B2B services, or another high-volume operational environment

Compensation

Salary: $110,000–$170,000 + performance bonuses & equity
Location: San Francisco, in-office
Schedule: Monday–Friday, in-office hours that match the rest of the company. The hours are long. The people who thrive here wouldn't have it any other way.

Benefits

Health, dental, and vision insurance

Learning & Knowledge Systems Lead