We're looking for an AI Product Manager who brings the full standard PM toolkit — user research, market and competitive analysis, roadmap and strategy, cross-functional delivery, and strong collaboration instincts - and applies it to products where the model underneath doesn't behave the same way twice. You have a real, hands-on feel for what different LLMs are actually good and bad at, and you use that to prototype ideas yourself, sometimes shipping small, production-quality AI features directly. You default to ownership: of the roadmap, of outcomes, of the quality bar that decides whether something's actually ready to ship, and of what an agent should be trusted to do on its own versus when a human needs to stay in the loop.

That's because building AI-native products means quality is a distribution, not a pass/fail — a feature can work correctly most of the time and still need a real answer for the failure tail, since the underlying system is non-deterministic, not just complex.

What you'll do

• Lead the roadmap and strategy for your product area, from problem discovery through delivery and post-launch iteration
• Lead the evaluation bar for the model-powered surfaces you're responsible for: define eval sets, identify failure modes, and know the rollback plan before a change ships
• Prototype product ideas directly using your own understanding of model strengths and weaknesses, and ship small, production-quality AI features yourself when that's the fastest path to learning
• Lead product decisions about agent autonomy: what the agent should be trusted to decide and act on independently, what needs a human in the loop, and how that line should move as trust in the system grows
• Treat prompting and context design as a product lever you use directly to shape behavior
• Partner with engineers on technical architecture with enough depth to challenge assumptions, propose alternatives, and influence design decisions
• Partner with Design and Research on UX so features are genuinely usable and understandable, not just technically correct
• Partner with Sales, Marketing, and Customer Success around releases — shaping GTM messaging, training, and rollout, and closing the loop on adoption signal afterward
• Track the metrics that actually matter for a non-deterministic system — quality/accuracy distribution, latency, cost-per-task — alongside the usual adoption and growth metrics

What you bring

• 2-5 years of product management experience owning a roadmap end-to-end, from strategy through shipped outcomes
• Direct experience shipping AI/LLM-powered features in production, including owning evaluation and quality decisions for them
• A real, hands-on feel for what different LLMs and model families are good and bad at, and the ability to prototype against that understanding rather than treating "AI" as a black box
• Real technical depth: comfortable in architecture discussions, and able to reason about tradeoffs (model choice, latency, cost, deterministic vs. LLM-based logic) as a peer to the engineers you work with
• Comfort defining and reasoning about eval sets and failure modes, and treating quality as a property of a non-deterministic system rather than a binary pass/fail
• Experience partnering cross-functionally with Sales, Marketing, and Customer Success around releases to connect product decisions to business growth
• A high ownership mindset: treating outcomes, not deliverables, as the responsibility
• You live in AI tools like Codex, Claude, OpenClaw, Cursor, or similar, and you've built your own personal agents or automated workflows with them
• Experience shipping features within strict privacy, security, and compliance constraints — treating what the system can retain, or send to a third party as a first-class product decision, not an afterthought.

Nice to have

• Experience with LLM observability/eval tooling (e.g., Braintrust, LangSmith)
• Background in a technical role (engineering, data science) before moving into product
• Experience in B2B, ideal if a market adjacent to UX/market research or insights/feedback
• Comfortable with how a lot of our AI features actually get built: fast prototypes/spikes that get real usage early, then get hardened directly into the product when they land, rather than a big-spec-first process

Of course, what is outlined above is an ideal set of expectations; however, business needs and other projects and tasks may shift, and additional tasks could be assigned at the discretion of your manager. If this role excites you but you're not sure you check every box, we'd still love to hear from you.

Note: Some of the benefits listed below apply only to U.S.-based employees. We offer a similarly competitive benefits program in the UK with many comparable offerings, which we’ll be happy to share with you during the interview process.