Head of Product, AI

At Productboard, we're building the first agentic product management system. We're rethinking how product management is done in a world full of AI agents, and how we help our customers build what matters for their audience. AI is changing what an interface even is: static screens are giving way to conversations, generated artifacts, rich interactive UI, and open-ended workflows, often on the same screen.

1. Spark, our agent. What it does, what it asks, when it acts on its own and when it waits for a person.
2. The agent harness and orchestration. Skills, scheduled tasks, human in the loop capability, model choices and cost per task. The runtime every agentic feature executes in.
3. AI experience. Spark UI, human in the loop UX and generative UI. How a person sees, steers and corrects agent work.
4. AI quality. You own the evals for Spark and every agentic feature: the eval suite, the judges, the method and tooling every team evals with, the review cadence that decides whether an output ships, and how evidence is shown to a person.
5. Feedback analysis and insights. The pipeline that turns customer conversations into structured insight, and the views where teams read it.
6. Product specs. Where the agent drafts with product teams, and what delivery ready means. The editor's storage and collaboration engine sit with the platform. You own the spec experience.
7. MCP servers. The surface through which Claude, Cursor and other agents work in Productboard. The tools, skills and agents we expose there. The protocol layer itself, auth, rate limits and hosting, sits with your platform peer.
8. Boards, roadmaps and dashboards. The views product teams work in. You own the application: what a board shows, how a roadmap reads, what a dashboard answers. The query engine underneath is your peer's platform capability.
9. Design for the whole product. All product designers report to you, on the AI product and on the platform. You own the design bar across the product. Critique, hiring and the design career ladder run through you. You also own the design system: the component library and design language every surface of the product uses, including the platform surfaces your peer owns.

1. Spark has a written quality bar, an eval suite that runs on every change, and a weekly review you chair.
2. Customer feedback attaches to the right capability and feature on its own, at a precision product teams trust. Executives read portfolio health on top of it.
3. For every place Spark shows up in the product, you have decided whether people use it through our UI, headless via MCP from an agent like Claude, or both.
4. Spark behaves the same way on every surface. One written document says when it acts, when it asks, how it shows its work, and how a person corrects it. Nothing new ships until critique has held it to that document.
5. The design team has a critique cadence, a hiring plan and a written ladder, run by you.
6. The Productboard MCP server is listed as a connector in the directories that matter: the Claude Connectors directory, the ChatGPT app directory, Grok, and the Cursor marketplace. MCP monetization runs on a clear, written strategy.
7. You know what share of Productboard work runs headless, through MCP and APIs, against what runs through our UI. You report it monthly and it is moving toward headless.

1. You know the product management job cold. You have run product management, or built products for product teams, and you understand discovery, prioritization, planning and roadmap communication from the inside. You know the use cases in detail: how feedback becomes insight, insight becomes a spec, and a spec becomes a roadmap, and where each step breaks today.
2. You have shipped an agent where the hard problem was trust. What the agent shows, when it asks, how a person reviews its work, and what happens when it is wrong. You can argue, with evidence, when an agent should act and when it should wait.
3. You run AI quality as a product discipline. You have owned evals for a shipped agent. You can read a trace, write a judge, and tell a model problem from a prompt problem from a product problem. You have held a release because the eval said no.
4. You have great UX DNA. You notice the experience problem before the designer flags it, and you have agent design work you can show. UX here means the product experience, not visual or brand design. You manage designers directly, hire them, and run critique, and you have contributed design work of your own that shipped, not only reviewed theirs.
5. You build with agents every day. You use Claude or its equivalent for real work, not to edit text. You have connected it to your own tools and data, built skills and multi step workflows that other people use, and you know where the model fails and how to fix the prompt, the context or the workflow.
6. You are fluent in headless. You understand that the product will be used through agents, via APIs and MCP, and you can decide which screens stay and which go.
7. You have a computer science or engineering background. You own the harness, the evals and the MCP surface, and those conversations go faster when you can read the code and the trace yourself.

• Your AI experience is model tuning or research. We need someone who has shipped agent behavior to customers.
• You are a platform PM. The platform has its own Head, and that search is open too.
• You are a brand or visual design leader. This role leads product design for the whole product, agent first.
• You cannot judge product design work yourself and rely on your designers to tell you what is good.