Almost every leadership team I speak with is discussing AI. That is understandable. The technology is moving quickly, the examples are visible, and nobody wants to discover too late that a competitor learned how to use it better.
But urgency creates a predictable risk: AI itself starts becoming the objective.
Teams begin with tools. They collect use cases. They launch pilots. They ask how many employees are using AI. Activity increases—and yet the business question sometimes remains surprisingly unclear.
Start with the business problem
Before discussing models, vendors or platforms, I prefer to ask what the organization is actually trying to improve. Is the constraint revenue? Cost? Cycle time? Quality? Customer experience? Decision speed? Risk? Management capacity?
The answer changes the AI conversation. A productivity tool, a forecasting system, a knowledge assistant and an automated decision process may all involve AI, but they create value in completely different ways.
This is where business judgment matters. The organization has to identify not only what is technically possible, but what is economically worth doing.
Value is more than a promising use case
A useful opportunity should survive several questions. How large is the problem? How frequently does it occur? What does the current process cost? What would improve if the solution worked? What new costs would appear? How much human oversight remains necessary?
Sometimes a glamorous AI opportunity loses its appeal when the economics are made visible. Sometimes a very ordinary use case—document review, internal knowledge retrieval, service triage, demand forecasting—creates far more value because it touches a high-volume process.
Adoption is part of the strategy
Even a strong technical solution can fail if people do not use it correctly. That is why I do not treat adoption as a communication exercise at the end of a project.
AI can change roles, workflows, decision rights, capability requirements and accountability. Leaders need to decide what people should still own, where human review is mandatory, what new skills are required, and how performance will be measured.
If those questions are ignored, organizations often end up with a pilot that technically works but never becomes a dependable operating capability.
Governance should enable—not merely restrict
Governance is sometimes introduced as the department that says no. I think that is too narrow. Good governance helps an organization move with greater confidence because the boundaries are clearer.
Privacy, security, reliability, bias, intellectual property, regulatory exposure and reputational risk should be considered in proportion to the decision being automated or supported. The right level of control for an internal drafting assistant is not the same as the right level of control for credit decisions, medical recommendations or employment decisions.
The executive discipline
For me, an AI strategy eventually comes down to disciplined choices: which problems deserve attention, which use cases deserve investment, which risks require control, which capabilities need to be built, and which experiments should be stopped.
That may sound less exciting than a technology showcase. But it is exactly where AI starts becoming a business capability rather than another wave of experimentation.
I share these reflections as practical starting points, not as a substitute for context-specific judgment. If your organization is working through a similar question, start a conversation with me →