AI projects often begin with a technology question: which model, platform or tool should we use?
But choosing the technology is rarely the most difficult decision. The harder questions concern how the organisation operates, who makes decisions and what happens when an AI system behaves unexpectedly.
The visible technology is only one part
An AI application depends on data, processes, infrastructure, governance and people. Each dependency introduces assumptions about how the wider organisation works.
A technically successful demonstration can therefore tell us surprisingly little about whether the system will deliver value in practice.
Decision rights matter
Consider three questions:
- Who decides whether an AI-generated recommendation should be acted upon?
- Who is accountable when the recommendation is wrong?
- Who has authority to change or suspend the system?
These are organisational questions, not merely technical ones.
From experiments to operational capability
The transition from pilot to production exposes dependencies that were easy to ignore during experimentation.
Teams need to understand how the system will be maintained, monitored and adapted as circumstances change.
The challenge is not simply making AI work. It is making AI work within an organisation that must remain accountable for its decisions.
What leaders should examine
Before committing to a major AI initiative, leaders should consider the assumptions on which its expected benefits depend.
That includes the organisation's ability to change its processes, resolve competing priorities and learn from what the technology reveals.
The technology matters. But it is only part of the system.