Four Gates: What AI Analysis Must Pass Before It Reaches a Decision

A leadership team looks at an AI-generated analysis on the screen. The first question is: “Where does this information come from?” The second: “Who stands behind this?” The third: “How was this reasoning arrived at?” Only when these questions get a fast, clear answer can the analysis make its way into the decision.

In their January 2026 Journal of Decision Systems article, researchers Saup, Asghar, Kanbach, and Kraus examined how a global multi-brand group has built a path from generative AI to strategic decision-making. Based on 27 leadership-level interviews, the study proposes four gates that an AI analysis must pass before leadership can accept it as part of a decision:

• Quality: Is the accuracy sufficient for a strategic decision?
• Provenance: Does anyone know where the data comes from?
• Explainability: Can the reasoning be opened up when needed?
• Accountability: Who stands behind the analysis?

These gates rest tightly on one another. Explainability depends on provenance, and accountability in turn depends on explainability. When all four conditions are met, the AI output travels to the leadership team as an unbroken chain: the data is traced, the logic is opened up, and accountability is named.

I would argue that half of building an AI strategy happens purely in the governance model. It is a matter of roles, operating models, and responsibilities. The other half concerns technological choices — investing in AI tools and pilot projects.

A named accountable owner, traceable provenance, and logic that can be explained out loud form a framework whose construction demands genuine leadership. In many organizations only one of these four gates is in place. The other three are still waiting for a leader who makes them mandatory.

Aspenly · Thinking