When AI initiatives are selected mainly because the technology is interesting, they often remain isolated pilots. When they are connected to genuine business priorities, they are more likely to attract ownership, resources, and sustained attention. The objective is not to have the greatest number of AI projects. It is to choose the initiatives that can create the greatest strategic value—and that the organisation is capable of implementing.

This requires leaders to answer two questions:

  1. Is this the right AI initiative for the business?
  2. Does the organization have the capacity to implement it?

The first question establishes strategic relevance, while the second takes us into culture, resources, and structure.

Three conditions determine whether AI ambitions become reality

In our work with strategy implementation, we repeatedly see three factors determine an organisation’s ability to deliver: culture, resources, and structure.

Imagine three companies introducing exactly the same AI solution. They may have similar technology and equally convincing business cases, yet achieve completely different results. The difference lies in what happens around the technology.

1. Culture: People adopt change when they understand it

Consider an insurance company introducing an AI assistant to help claims handlers review cases. The pilot is promising: the tool retrieves relevant information, summarises documentation, and proposes a next step. Management expects shorter processing times and more consistent decisions.

But employees hear a different message. They wonder whether the system is being introduced to reduce headcount. Experienced claims handlers feel that their professional judgment is being devalued. Some distrust the recommendations, others fear being blamed if they follow one that turns out to be wrong. Team leaders, under pressure to deliver immediate efficiency gains, begin comparing adoption rates.

The result

Employees attend the training, then quietly return to their old ways of working. Some use the tool only when a manager is watching. Others copy its output into their work without properly evaluating it. On paper, the technology has been implemented, but in practice, very little has changed. New risks may have even been introduced.

A better approach

Leadership explains that the objective is to remove repetitive work and give claims handlers more time for complex cases and customer dialogue. Experienced employees help design the new workflow and test when the tool should, and should not be trusted. Managers make it clear that human judgment remains essential. Teams openly discuss errors, limitations, and improvements without turning every mistake into a search for someone to blame.

The technology is the same but the cultural conditions are different.

Ask yourselves:
  • Do employees understand why AI is being introduced?
  • Do they believe it will help them perform their work—or threaten their position?
  • Are people comfortable challenging the system’s output?
  • Do leaders demonstrate the new behaviour themselves?
  • Is experimentation encouraged while accountability remains clear?

AI adoption depends on trust, involvement, and psychological safety. As with any change in strategy, if people do not understand or believe in the change, even excellent technology can become an expensive icon nobody clicks.

2. Resources: Progress follows when priorities are properly resourced

Now consider a retailer that wants to use AI to improve demand forecasting. The potential value is substantial: fewer stockouts, less waste, and better working-capital management. The initiative is declared a strategic priority. A steering group is formed, a technology partner is selected, and an ambitious launch date is announced.

But the people needed to make it work already have full-time responsibilities. The data team is committed to several other transformation projects. Store managers are asked to validate forecasts during their busiest trading period. The commercial team has no capacity to redesign purchasing routines. Historical data is fragmented across different systems, but no budget has been allocated to clean it.

The result

Everyone agrees that the project matters but nobody has the time to deliver it. Deadlines begin to slip. The external partner produces a technically sophisticated model, but the business does not trust the data behind it. Managers continue ordering products through spreadsheets and experience. After several months, the initiative is described as a pilot that “did not demonstrate sufficient value.”

The real problem was not the AI at all. The organisation attempted to add a strategic transformation on top of business as usual, without deciding what would receive less attention.

A better approach

Start with an honest capacity assessment. The retailer selects two product categories rather than attempting an immediate company-wide rollout. It assigns a dedicated product owner, protects time from data and commercial specialists, and funds the necessary data preparation. Store teams participate at defined stages instead of being continuously pulled into the project. Other lower-priority initiatives are paused.

Ask yourselves:
  • Have we assigned people or merely placed their names on a project plan?
  • What existing work will stop or receive less attention?
  • Is the necessary data available and reliable?
  • Do we have the relevant business, technical, and change capabilities?
  • Is the investment sufficient to move beyond a demonstration?

Simply calling something a priority does not automatically create capacity. If people, time, data, and investment are not reallocated, an AI ambition remains an additional task competing with everything the organisation was already doing.

3. Structure: Clear ownership helps good ideas move forward

Lastly, consider a business-to-business company introducing an AI tool to help sales teams prepare customer proposals.

Sales believes IT owns the initiative because it involves technology. IT believes Sales owns it because the value must come from changed commercial behaviour. Marketing wants control over the generated content. Legal is concerned about confidential data and unsubstantiated claims. HR assumes training belongs with the project team but the project team is made up of people contributing alongside their regular jobs. Everyone is involved but nobody is accountable.

The result

The pilot produces encouraging feedback, but important decisions remain unresolved. Which customer data may be used? Who approves the generated content? Should the tool connect to the CRM system? How will proposal quality be measured? Who has authority to change the sales process? Meetings multiply, but progress slows. The pilot remains a pilot—not because it failed, but because the organisation never created a mechanism for turning it into normal operations.

A better approach

In the stronger version, one commercial executive is accountable for the business outcome. A named product owner manages the implementation day to day. Sales, IT, Marketing, and Legal have clearly defined decision rights. The team agrees on a small number of measures: preparation time, proposal quality, usage, conversion, and identified risks. Progress is reviewed every two weeks, with a clear route for escalating decisions.

The tool is introduced to one sales unit first. Feedback is used to improve the workflow before the next rollout. Ownership gradually moves from the project team into the commercial organisation, where the new process must ultimately live.

Ask yourselves:
  • Who is accountable for the business result?
  • Who can make day-to-day decisions?
  • Which functions must contribute, and what authority does each have?
  • How will progress, value, adoption, and risk be measured?
  • How will the initiative move from pilot to normal operations?

Structure does not mean creating more meetings or a large central AI committee. It means creating clarity: clear ownership, clear decisions, clear measures, and a consistent rhythm for learning and action.

The same technology, three different realities

These examples reveal why AI implementation cannot be delegated to the technology function alone. A company can purchase the right tool and still fail because employees do not trust it. It can choose a valuable use case and still fail because nobody has been given the capacity to implement it. It can run a successful pilot and still fail because ownership and decision-making remain unclear.

Before launching another AI initiative, leaders should therefore ask four practical questions:

  • Business relevance: Are we solving an important business problem?
  • Culture: Will our people understand, trust, and adopt the change?
  • Resources: Have we provided the real capacity needed to deliver it?
  • Structure: Is it completely clear who owns the result and how progress will be managed?

If the first answer is uncertain, the organisation may be pursuing technology without a strategic purpose. If any of the other three answers is uncertain, it is not yet facing a technology problem. It is facing an implementation problem.

‍