Domain Expert, Insurance

Bring deep insurance expertise into the room where enterprise agents get designed, and turn it into specifications engineers can build the same week. About Sycamore Sycamore is building the trusted agent operating system for the enterprise. Our platform helps companies build, deploy, and orchestrate AI agents that take on real operational work, with the security and control large organizations need. We are a small, engineering-led team working directly with Fortune 500 enterprises. We have raised $65M from Coatue and Lightspeed, along with other investors and industry leaders. The role The hardest part of deploying an agent into an enterprise workflow is almost never the model. It is knowing what the workflow actually is: who decides what, under whose authority, against which rules, with what evidence, and what happens when the answer is wrong. That knowledge lives in people who have done the work. You are one of them. You will pair directly with our Applied AI engineers on customer deployments. They build; you supply the substance that makes what they build correct. Your output is not advice or a slide. It is a specification precise enough to become a workflow definition, a rule set, a review screen, and a set of thresholds. You will then judge whether what came back is actually right, which is a different question from whether it runs. Our platform is built around this shape of work. It extracts structured facts from documents with citations back to the source, applies deterministic rules where rules exist, routes the ambiguous middle to a human with the right permissions and authority limit, and keeps a tamper-evident record of every decision. Each of those pieces needs someone who knows what correct looks like. You will also be in front of customers throughout: discovery sessions, requirements work, design review, and the conversations where a skeptical operator decides whether we understand their job. Credibility with those people is a large part of the role. Insurance at Sycamore Insurance runs on documents and judgment at scale, and the volumes are enormous. Submission intake and triage. Underwriting review and referral. First notice of loss through adjudication and settlement. Policy servicing and endorsements. Subrogation and recovery. Complaints and regulatory inquiries. Much of this moves through email, PDFs, and policy administration systems that will not be replaced. The decisions need to be consistent, explainable, and defensible, to a regulator, to a reinsurer, and to the policyholder. That makes it an unusually good fit for our platform and an unusually bad fit for anything that cannot show its work. Agents need to operate inside authority limits, elevate to a licensed human at the right threshold, and record why a recommendation was made. What the work looks like In one month, you might: Assemble the state-by-state rule corpus a claims or underwriting workflow must apply, and mark which rules must never be left to agent judgment. Specify adjuster or underwriter authority limits and the escalation ladder that applies when a case exceeds them. Define what leakage actually means for a given line of business, so the system is measured against the number the business manages. Write down what a Department of Insurance examination would ask of this process, and what evidence would answer it. Sit through a discovery session and leave with the process written down properly: decision points, authority limits, service levels, exception paths, and what happens on breach. Work with an engineer to decide which parts of a workflow should be deterministic rules and which genuinely require judgment, and be honest about which is which. Specify the review screen: what a reviewer must see to decide in thirty seconds, and which fields are the ones people actually get wrong. Set the thresholds for what is handled automatically, what is routed to a person, and what is refused outright, along with the relative cost of a false positive and a false negative. Review what an agent produced across a batch of real cases and tell the team where it is wrong, not just where it failed. Identify which system of record a workflow depends on and what integrating with it honestly requires, including what will still be manual. What you will do Serve as the domain authority on customer deployments in your industry, from discovery through production. Translate real operational processes into specifications precise enough to build from: decision points, roles and authority, rules, thresholds, escalation, and evidence requirements. Build trust with customer operators, executives, and control functions as someone who has done their job. Feed patterns back to Product and Core AI so that what we learn in one deployment becomes a platform capability rather than a repeated conversation. Help our commercial team evaluate which opportunities in your industry are real and which are not. What we are looking for 10+ years in insurance, in a role where you owned or governed the actual process: claims, underwriting, policy operations, compliance, or product. Deep familiarity with the operating model: authority limits and referral thresholds, fair claims handling standards, state-by-state regulatory variation, and the legacy policy administration estate. Direct experience with the operational reality of at least one of: claims handling, underwriting and submission triage, policy servicing, or regulatory complaint handling. The ability to write a process down precisely. If you can explain a workflow’s decision points, authority limits, and exception paths clearly enough that an engineer can build it, that is the core skill. Judgment about automation. You know which parts of your domain are genuinely rule-based, which require expertise, and which should never be delegated to software at all. Executive presence and high EQ. You can be the credible voice with a customer’s operators and their leadership in the same conversation. Intellectual honesty. You will sometimes need to tell us that something we built is wrong, or that a customer’s request is a bad idea. AI-native. You use modern AI systems in your own work and can reason about where they help and where they mislead. Comfort with startup ambiguity, broad ownership, and regular customer travel. You do not need to be technical in the software sense. You do need to be precise, and you need to be genuinely curious about how the system works. A 30-minute introductory conversation. Two 60-minute conversations, one on the operational depth of your domain and one on how you would specify a workflow for a build team. A working session where you take a real process from your industry, specify it end to end, and defend the decisions you made about authority, thresholds, and what should not be automated. Put deep domain expertise directly into production software rather than into recommendations. Work alongside engineers who will build what you specify, in the same week. Define how agents handle consequential work in an industry you know well. Sit at the front of a category that is still being defined, with real influence over what gets built. Join early enough to shape how Sycamore approaches your industry. Receive competitive cash compensation and meaningful equity in the company you are helping build. Hard problems, real impact Trust architectures, memory systems, multi-agent coordination. The foundational layer that makes AI agents work in production. Small team, high ownership Every engineer shapes the product and the culture. No layers of process between you and the work that matters. Backed by the best $65M from Coatue, Lightspeed, Abstract Ventures, Dell Technologies Capital, 8VC, and notable industry angels. Grow with us Competitive compensation, meaningful equity, and a genuine focus on your growth as the company scales.

• Job Posting: Domain Expert, InsuranceApplied AI · Palo Alto · full-time
• J-18808-Ljbffr Sycamore