CDO Vision New York 2026 | Panel Discussion: The Boardroom Conversation - How CXOs Shape AI Strategy

AI adoption at the enterprise level is no longer limited by ambition or access to technology. Boards are approving investments, pilots are moving into production, and AI systems are being built into core operations. What remains unresolved is how organisations retain the reasoning behind these decisions once execution begins, particularly as responsibility shifts from strategy to delivery.

At CDO Vision New York 2026, a panel discussion titled The Boardroom Conversation on How CXOs Shape AI Strategy examined how AI decisions are taken at the executive level and what happens to those decisions once implementation begins. The discussion drew on perspectives from Tsvi Gal of Memorial Sloan Kettering Cancer Center, Douglas Hegley of The Metropolitan Museum of Art, Mikhail Lisovich of Redwood Logistics, David Fapohunda of Barclays, and Henry Tirri of Nokia, and focused on approvals, accountability, and why organisations often struggle to explain the reasoning behind AI choices after time has passed.

As AI programmes scale, decision-making increasingly happens under compressed timelines and incomplete information, with trade-offs accepted on the assumption that refinement can follow. This approach enables organisations to move faster, but it also introduces a structural weakness. When outcomes are reviewed months later, teams often find it difficult to explain not what was implemented, but why the decision was considered sound at the time.

AI amplifies this challenge because its effects rarely remain contained within a single function. Once systems move into production, attention shifts quickly to performance metrics and delivery milestones. In that transition, the logic that connected the original business problem to the chosen solution is often lost.

Why context fades once execution begins

In many organisations, executive decisions rely heavily on shared understanding rather than durable records. Approvals typically occur in meetings, supported by presentations, assumptions, and verbal alignment, with the expectation that collective memory will carry intent forward into execution. That approach works when conditions remain stable and leadership teams remain unchanged.

The weakness becomes visible when scrutiny increases, such as during regulatory reviews that require formal justification or board discussions that revisit earlier assumptions. Leadership transitions can expose the same gap, particularly when new executives inherit AI systems they did not approve. In these moments, organisations are forced to reconstruct intent after the fact, often relying on fragmented documentation that captures outcomes but not judgment.

Most AI governance frameworks are designed to manage control, compliance, and risk thresholds. They define what is permitted, establish oversight mechanisms, and monitor system performance. What they tend to preserve less effectively is the reasoning behind decisions, including why a certain level of uncertainty was acceptable or why deploying an imperfect system was preferable to delaying action.

As AI becomes more visible at the board level, expectations rise accordingly. There is growing pressure to demonstrate progress, show momentum, and avoid falling behind peers. This encourages execution while pushing explanation into the background. Over time, that imbalance weakens confidence, even when systems perform as expected, because the organisation struggles to defend its choices with clarity.

Treating decisions as durable assets

One response discussed during the session was the need to treat decisions themselves as part of the enterprise asset base. This involves capturing not only approvals, but also intent, constraints, and expected outcomes at the point when choices are made.

When AI initiatives are tied to clearly defined outcomes from the outset, evaluation becomes more disciplined. Financial impact, operational improvement, risk mitigation, or mission outcomes are articulated early, and review points are established beyond routine reporting cycles. The reasoning behind selecting one approach over another remains connected to the decision as it moves through the organisation.

This shift changes how accountability functions. Outcomes are no longer assessed in isolation from the conditions under which decisions were made. Instead, organisations can show how judgment evolved as circumstances changed, with adjustments documented as part of an ongoing process rather than framed as corrective action.

In regulated and public-facing environments, this discipline carries particular importance. AI decisions increasingly intersect with external obligations to customers, regulators, and shareholders. When reasoning is traceable, organisations can demonstrate that choices were made deliberately, with an understanding of trade-offs, rather than deferred to technology or urgency.

AI systems will continue to change as data, expectations, and operating environments evolve. What should remain stable is the organisation’s ability to explain itself. Preserving decision context is not an administrative exercise, but a governance capability. As AI becomes embedded in enterprise strategy, that capability determines how effectively organisations maintain trust, accountability, and control over the systems they deploy.