Four signs your AI rollout is solving the wrong problem

The first frown appears when the AI rollout is handed to the CTO. The second arrives when a newly appointed Chief AI Officer is asked to set the strategy alone. The third comes while I am looking at a dashboard that treats app openings and prompt length as proof of adoption. The fourth appears when somebody announces that the organisation has chosen its main AI tool, as if selecting a vendor has settled the strategic question.

Each decision can look sensible in isolation. Technology needs an owner. A senior AI appointment can create focus. Usage data has a place. Enterprise procurement is unavoidable. The trouble begins when any one of them is asked to carry work it cannot do.

AI rollout is a behaviour-change programme with a technical layer.

Frown one: the CTO owns the rollout

The CTO keeps the environment stable, the data moving and the security risk within acceptable bounds. That role is essential. It does not automatically make the technology function responsible for changing how thousands of people judge, decide, collaborate and learn.

Asking the CTO to own all of that is a category error. It is like asking the plumber to redesign the kitchen because both jobs involve water. The technical architecture and the way people use it are connected, but they require different expertise and different forms of authority.

AI adoption needs shared accountability. Technology owns the environment and integration. Legal and Compliance define boundaries. Business leaders own value and the redesign of work. HR and L&D own much of the capability, leadership behaviour and cultural permission around the change. One executive may coordinate the programme, but no single function can substitute for the others.

The practical question is not who owns AI. Ask who owns each decision after the launch: who approves a use case, who protects experimentation time, who reviews a high-consequence output and who decides whether a workflow should change.

Frown two: the new AI leader writes the strategy

A capable Chief AI Officer can connect work that is scattered across the organisation. The title does not create lived context overnight. A leader who arrived three months ago cannot yet know where customers feel friction, which processes survive only through workarounds or why the last transformation stalled in one particular function.

The people who understand where AI may create or destroy value are already on the payroll. Some have been close to the work for years. Others are junior enough to see an absurd task before the organisation teaches them to accept it. Most have never been asked what they know in a way that can shape strategy.

The AI leader’s first job is therefore discovery. Find the employees already experimenting, the managers who know where work gets stuck and the informal bridges who carry trusted advice between functions. Bring those people into the strategy before presenting it back to them.

A strategy gains credibility when people can recognise their work inside it. Without that context, even an intelligent plan becomes corporate theatre: polished, visible and strangely detached from the organisation expected to carry it out.

Frown three: activity is reported as adoption

Tracking adoption through app openings is the corporate equivalent of measuring a gym membership by how often somebody swipes in. The person can arrive every morning and still be in worse shape by June. Presence shows that access exists. It says little about what happened after entry.

Prompt length is not much better. A long prompt may reflect a sophisticated workflow or a confused person wrestling with the wrong tool. Neither interpretation tells a leader whether the organisation improved a decision, removed waste or created something it could not produce before.

Usage metrics are useful as operational signals. They can show whether access is broken, whether a launch reached the intended group and where support may be needed. They become vanity metrics when they are presented as business outcomes.

• Which decision changed because better evidence became available?
• Which hours were recovered, and where were they reinvested?
• Which process was redesigned or removed?
• What useful work exists now that did not exist last quarter?
• Where did quality improve without increasing risk or rework?

Those measures are harder to collect because they sit close to the work. That is precisely why they matter. A dashboard should tell a leader whether practice changed, not simply whether software was opened.

Frown four: the main tool becomes the strategy

Choosing an enterprise AI platform is a procurement and architecture decision. It can simplify access, support governance and give employees a common starting point. It does not answer which problems deserve attention or how the organisation will keep learning as capabilities change.

A single approved environment may be right for sensitive work. It should not become an intellectual boundary around what leaders are allowed to understand. The frontier moves too quickly for any organisation to confuse today’s contract with tomorrow’s strategy.

The answer is not an ungoverned collection of fashionable tools. Build a clear core environment for ordinary work, then create controlled routes for testing other capabilities against specific questions. Keep the task and the decision stable while the tool choices evolve around them.

This also protects the organisation from vendor-shaped thinking. When the licensed platform defines the use cases, the programme begins to solve the problems the product happens to display well. Strategy should move in the other direction: begin with the work, then select the capability that fits it.

Turn the four frowns into four decisions

Shared ownership replaces the assumption that the CTO can carry the change alone. Organisational discovery gives the AI leader context before strategy hardens. Outcome measures replace activity as the definition of adoption. A governed portfolio keeps procurement in service of the work.

Organisations that reverse those choices usually get the familiar result: a dashboard, a vendor and a small population of frustrated people whose capability has moved faster than the institution around them. The sharpest eventually leave for places where they can use what they know.

If your company is early in its AI work, this is the useful moment to correct course. Decide who owns the human change. Ask the people who understand the work. Measure what became different. Keep the strategy larger than the tool selected to support it.