Organizations implementing responsible agentic AI must institute life-cycle governance: assigning clear human accountability and responsibility as well as mapping risks, measuring them, and managing them throughout design, development, deployment, and monitoring.

Human Accountability and Responsibility

Builders of responsible high-stakes agentic AI must implement clear ownership and decisionmaking pathways before any deployment. We recommend that every team include five roles: an accountable owner, an evaluation lead, a security lead, a transparency lead, and a responsible agentic AI lead.

In smaller or pilot deployments, the accountable owner may wear several or all of these hats. But for high-stakes policy and research use, these roles should ideally be held by separate individuals. In some cases, each role may involve multiple individuals with the expertise indicated.

Staged Rollout with Clear Processes

Building and architecting agentic AI processes should follow best practices in the technology industry:

• Tools should be built in stages and continuously improved. Each stage should have clear goals, iterate toward them on a regular cadence driven by the accountable owner, and create artifacts that meet internal and user needs but minimize waste.
• Part of that process should include clear phases of development with built-in requirements and transparency artifacts that are subject to review and approval.
• Organizations should name a responsible agentic AI lead that oversees and governs approvals to move to each stage internally. That lead may be responsible for collecting input from important stakeholders across the organization—such as the technology, finance, and legal departments—to approve the move to the next phase.
• In addition, the responsible agentic AI lead should ensure that ongoing monitoring, review, and mitigation are completed as required.
• The responsible agentic AI lead should be empowered to reject or pause agents that do not meet the organization’s criteria at each phase, ensuring that the organization only approves secure, transparent, useful, well-governed, and responsible agents that show clear return on investment and risk.

This process should ensure that accountable owners are truly accountable to the requirements in this playbook.

Next section: Principle 3: Minimize Known Risks