Learning & Knowledge Systems Lead

About the role

This role sits at the intersection of operations, engineering, and revenue operations inside an AI-native commercial insurance company that is rebuilding the entire business as software. You will turn the judgment locked inside Harper's best operators into AI-legible knowledge by creating living documents, decision logs, and retrievable skills that agents can actually call. The role exists because AI cannot understand a company by default and only works when business processes are documented clearly enough for systems to retrieve the right context, recognize workflows, handle edge cases, and escalate when human judgment is required. You will partner closely with operators across sales, intake, service, placements, and renewals to extract how they actually think and translate that into structured, usable knowledge. This is not a corporate learning and development position; there will be no LMS, slide decks, or e-learning projects, only knowledge engineering that makes human expertise machine-actionable. You will be expected to move at the same fast pace as the company and to use advanced AI tools such as Cursor, Claude Code, MCP servers, agent memory files, and structured prompting to deliver durable systems.

Key facts

What you'll do

• Embed with sales, intake, service, placements, and renewals operators to sit in on workflows, listen to calls, and shadow real work in order to document how experts actually operate.
• Convert transcripts, Slack threads, Looms, and one-off explanations into source-of-truth documents, decision logs, playbooks, process maps, glossaries, and system-boundary notes that clarify where systems start and stop.
• Turn stabilized rules and expert judgment into skills and functions that AI agents can call, and collaborate with engineering on refresh automations so documentation stays alive instead of going stale.
• Hunt for edge cases such as reworks, escalations, stale quotes, market follow-ups, binder and payment gaps, and customer confusion, and capture them before they recur.
• Distill AI-generated playbooks into executable rollout plans with named owners, the first three moves, a rollout cadence, and specific dates so that plans actually run.
• Build onboarding paths and setup scripts that get new hires operational in Cursor, Claude Code, and the harness within their first week on the job.
• Run cohort rollouts, drive adoption of knowledge systems, and make activity visible so the company can see who is actively using the harness and acting on documented skills.
• Shape meetings in real time so they produce useful artifacts including decisions, owners, definitions, edge cases, open questions, and next steps that feed directly into knowledge systems.
• When the same problem appears multiple times, convert it into a playbook, a QA check, a skill, or a product requirement and escalate to the CEO when extraction demands it.
• Partner with product and engineering to turn captured knowledge into requirements, feature flags, and guardrails that prevent recurring failures.
• Maintain a clear line of sight between frontline operator behavior and back-office systems so that AI models always retrieve the freshest and most relevant context.
• Continuously measure the quality and usage of knowledge assets, iterate on structure and clarity, and remove ambiguity that causes rework or hallucination.

Requirements

• An exceptional writer and synthesizer who can take a messy transcript and turn it into a clear operating document, and then turn that document into something a team actually executes against.
• AI-native in practice, meaning you have a trained sense for when an output is structurally wrong and not just stylistically off, and you prompt for extraction of decisions, contradictions, owners, and edge cases rather than only summarization.
• Comfortable using tools such as Cursor, Claude Code, MCP servers, agent memory files, and structured prompting to build and maintain knowledge systems.
• Genuinely curious about how organizations work and happy to sit with operators to discover how work actually happens, not how it is supposed to happen on a diagram.
• Structured but not bureaucratic, caring more about whether documentation changes behavior than about how polished a deck looks.
• Able to operate comfortably in chaos without becoming chaotic, staying low-ego, persistent, and allergic to the mindset that someone should probably document that later.
• Backgrounds that can work include modern 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, or AI ops / human-in-the-loop systems.
• Comfortable moving quickly in a high-growth environment with long days, on-site presence, and very high standards for execution and judgment.

Practical notes

LENGTH: 700-900 words. No HTML, no markdown, no em dashes.

Output the page only.