Your job description still says "product manager." Your actual job changed six months ago.
Harvard Business Review named the shift in February: companies need "agent managers" — people accountable for what autonomous systems do, not just what gets built. The finding that stung: domain expertise matters more than AI expertise. The best agent managers come from roles where they already understand the business process being automated.
That's you. The PM who knows why the approval flow works the way it does. Who remembers the edge case that crashed onboarding in Q3. Who can explain, in plain language, what "done" looks like for a customer support interaction.
Except nobody told you the job changed. And nobody gave you the tools.
The shift nobody prepared for
Deloitte's 2026 State of AI in the Enterprise survey (3,235 leaders, 24 countries) found that 34% of organisations are now using AI to deeply transform — reinventing core processes or creating entirely new products. Another 30% are redesigning key processes around AI. Close to 75% of businesses plan to deploy AI agents by the end of this year.
Meanwhile, Productboard's CPO survey found 85% of leaders are investing in AI tools. Only 2% consider talent development their biggest priority.
Read that gap again. Three-quarters of companies deploying agents. Two percent prioritising the people who'll govern them.
The product manager sits directly in that gap. You're the person who translates business intent into system behaviour. That translation used to produce feature specs and user stories. Now it produces rules — the constraints, guardrails, and governance logic that determine what agents can and cannot do.
From "what should we build" to "what must agents never do"
Raconteur's 2026 governance report makes the distinction sharp: existing governance frameworks are not designed for autonomous decision-making. Agents that act independently need someone defining their boundaries before they act, not cleaning up after.
That someone has always been the product manager — just under a different name. You've always written acceptance criteria. You've always defined the "must nots." You've always been the person who says "no, that edge case breaks everything, handle it before shipping."
The difference: your acceptance criteria used to govern developers who could ask clarifying questions. Now they govern agents that will interpret literally, act independently, and scale mistakes instantly.
Productboard's 2026 survey confirms the instinct: 59% of PMs say strategy and business acumen are the most important skills for the next two to three years. Not prompt engineering. Not model architecture. Strategy — knowing what the system should do. Business acumen — knowing why.
The gap is the opportunity
The consulting firms are writing frameworks for CISOs and compliance officers. The AI companies are building orchestration layers for engineering teams. Nobody is equipping the PM — the person closest to the business logic, the customer context, and the workflow intent — with tools to design rules that agents actually follow.
The PM's new job isn't learning to code agents. It's learning to govern them. To write the rules that fire. To encode the business intent that no agent should violate. To be the "rule designer" — the person who turns institutional knowledge into machine-readable constraints.
Your domain expertise isn't obsolete. It's the single most valuable input the agent needs. The only question is whether anyone gives you the tool to encode it.
Sources: Harvard Business Review, "To Thrive in the AI Era, Companies Need Agent Managers" (February 2026); Deloitte State of AI in the Enterprise 2026 (3,235 leaders, 24 countries); Raconteur, "Autonomous AI agents 2026: the new rules for business governance" (2026); Productboard AI Survey 2026.