132 M C K I N S EY Q UA RT E R LY industries. Directors ask management to map the elements, including vendor partnership, cap ital allocation, and workforce readiness, it would require to catch up quickly if needed. Boardroom test questions include the following: - How are we tracking AI developments inside the industry, as well as with competitors, to deter mine what actions we should take? - Do we have a credible plan to follow fast on a proven AI capability, and how do we assess our readiness to move? What are the risks associated with not pivoting in time in various business areas? SIX ACTIONS TO TAKE Our research highlights six actions that boards should consider taking, with the degree of pursuit varying per their AI posture: Board members should become comfortable with AI by using it in their personal lives. 1. Align on AI posture and review it annually (at least). The very first order of business for directors is to align on what posture the organization should take with AI—without that clarity, none of the other actions matter. Boards should then regularly revisit their AI posture in response to changes in the competitive, regulatory, and technological environ ments. Proactive posture reviews are a useful way to make sure that the company’s stance reflects today’s realities rather than yesterday’s assumptions. Of course, this annual review shouldn’t replace more frequent engagement on the topic (see action number four). - 2. Clarify ownership of AI over- sight—within both the board and management. Oversight fails without clear accountability. Boards should explicitly define which topics should be reviewed and fully discussed in full- board sessions (for example, material investments to scale enterprise-wide AI), which belong in committees (for example, risk frameworks and material vendor reviews), and which do not require significant board discussion (such as regular operational decisions). Without this specificity, ambiguity emerges and accountability breaks down or precious agenda time is wasted. 3. Codify a framework for AI governance policy. Most companies draft principles or ethics state ments, but fewer than 25 percent of companies have board-approved, structured AI policies. A credible governance framework designed to last should specify the following: - scaling rules (when pilots earn capital to scale enterprise-wide) risk thresholds (where human sign-off is neces sary and what guardrails should be in place) - vendor or data guardrails (IP protections, third- party audit rights, security, and lineage standards) escalation triggers (what incidents reach the board and how fast) 4. Engage more broadly (and more frequently) with those doing the work. It is not enough to engage only with CEOs or CFOs on AI developments across the organization. Board directors should be regularly exposed to and interact directly with the executives who are embedding AI into operations (such as chief data and analytics officers and business and division leaders) to gain a deeper under standing of progress against goals and impact on competitive dynamics. - 5. Tie AI investment to business value. Boards should encourage management to not only identify but also quantify the potential opportuni ties and the possible risks associated with AI adoption. This view can help boards guide businesses through the short- and long-term trade-offs that balance opportunity and risk, using their AI posture. Effectively - An online version of this article is available on McKinsey.com AI Reckoning