Senior Product Manager - AI

Our product roadmap increasingly depends on AI doing real work inside the product: extracting structured data from documents and running agents that take action on a user's behalf. That only works if two things are true: the underlying extraction is accurate enough to trust, and every AI action is governed (auditable, explainable, escalated to a human when it should be, and safely isolated between customers). This role owns that platform. You turn the AI isn't accurate enough yet, and agents need guardrails into a roadmap, a trust bar, and an interface that other product teams can build against without a standing dependency on one engineer.

The core responsibilities for the job include the following:

Extraction accuracy:

• Own the roadmap to improve document/data extraction accuracy against defined targets, including major build-vs-improve calls on the underlying approach, in partnership with ML/engineering.
• Define what "accurate enough to trust" means for each use case, and track progress against it on a live, shared metric.

Agent runtime and governance:

• Own product requirements for the runtime that lets AI agents take actions inside the product.
• Own the governance layer: audit trail, explainability, human-in-the-loop escalation, policy enforcement, and tenant isolation. This is what makes an AI-driven action safe to ship, not just possible to build.
• Define a policy-bounded interface for agent-triggered actions, so actions stay within defined guardrails without requiring engineering review per action.

Platform-as-a-service discipline:

• Define and maintain a self-serve interface so other product teams can build on this platform without filing a ticket to a single engineer for routine requests.
• Partner with adjacent platform owners (e. g., core integrations/data) so teams building on top see one coherent surface, not competing conventions.
• Push back on one-off requests that would turn the platform back into a bottleneck.

What success looks like (first 2-3 quarters):

• A clear, evidence-based call made on the extraction approach, with a dated accuracy trajectory behind it rather than opinion.
• Defined accuracy targets met on priority use cases, tracked on a live dashboard.
• At least one AI-driven agent live in production with the escalation/governance loop actually working, not just the underlying model.
• At least one other product team consuming the platform self-serve, with no standing ticket queue for routine requests.
• Usage and override rates on agent-triggered actions instrumented and reviewed.

Requirements:

• 4+ years in product management, including a stretch owning a platform, API, or infrastructure-facing product used by other internal teams, not exclusively customer-facing feature work.
• Direct experience shipping ML/AI-backed product where trust and accuracy were the product, not a feature bullet. You've defined a precision/recall bar and held engineering to it.
• Working fluency in what governance means for an AI system in production: audit trails, explainability, human-in-the-loop design, policy enforcement, and tenant/data isolation. You don't need to have built all of these, but you've shipped alongside at least one and know why each exists.
• Comfortable owning a roadmap with genuine capability uncertainty and communicating that uncertainty honestly to stakeholders instead of smoothing it over.
• Platform-PM instincts: default to how do I make this self-serve over how do I build this feature for one team. Says no to scope creep that would make you a bottleneck.
• Senior-level autonomy: given a problem area rather than a spec, you set the why, frame the what and how, and are trusted with the area's roadmap and prioritization end to end.

Nice to have:

• Prior experience in a regulated domain where explainability/audit was a hard requirement rather than a nice-to-have.
• Exposure to agent-action protocols (e. g., MCP) or policy-engine design for agentic systems.

Bonus points:

• None of these are required, but they will set you apart:
• CLM, legal-tech, or contracts domain experience.
• Direct customer-facing experience in a sales, customer success, solutions, or pre-sales role or a track record of close partnership with those teams.
• Experience shipping AI-native or AI-assisted products to production.
• A background in a fast-scaling B2B SaaS environment.