At a recent event on AI in the workplace, one speaker said: “Giving everyone an AI subscription is a losing strategy.” I couldn’t agree more. In fact, the number of AI licences your organisation has bought tells you very little about the value AI creates.

In my work, I see the same pattern again and again. Organisations report how many licences they have purchased, how many tools they have deployed and how many employees have completed AI training. These numbers look good in a management report. However, they say very little about whether AI makes the business faster, better or more profitable. Access to AI does not automatically create value. And even when people use AI, that does not mean they trust it or use it well.

 

The problem: we measure what is easy to count

 

Licences, tool rollouts and training completions are easy to count, so they end up in the reports. But when I look closer at organisations that have rolled out AI broadly, I usually find a very different picture.

 

What this looks like in practice

• Usage drops after the first few weeks. In the weeks after the rollout, many employees try the new tool out of curiosity. After a while, however, daily pressure takes over, and people go back to their usual way of working because AI is not yet part of their tasks.
• People use AI mainly for simple tasks. Most employees use AI to write emails, summarise documents or translate texts. These tasks save a few minutes, but they rarely touch the processes that drive revenue, quality or customer satisfaction.
• People either trust AI too much or not at all. Some employees copy AI results without checking them, which creates errors and risk. Others, in contrast, do not trust the output and check every detail twice, so they lose the time they hoped to save.
• Managers cannot name the impact. When you ask managers what has changed in their team’s results since the rollout, they often struggle to answer. They know that people use AI, but they cannot point to faster processes, better quality or lower costs.
• Saved time cannot be traced. Employee surveys often report that AI saves several hours per week. Yet nobody can show where those hours went, because no one decided in advance how to use them.

 

When leaders see these signs, they often conclude that people need more training and buy another round of courses. As a result, the underlying causes stay the same, and so do the results.

 

 

 

Why this happens

 

From my experience, there are four main reasons.

Leaders treat trust as a communication task: Leadership announces that the organisation has approved a tool and that it is safe. That reassurance helps, but people decide whether they trust AI while they use it, task by task. So if they cannot understand or question a result, a statement from the top will not change their behaviour.

Each function looks at AI from its own angle: IT focuses on tools and security. Legal and compliance focus on risk, while HR focuses on training. Meanwhile, the business focuses on results. Each of them does good work, but nobody brings these views together for a single use case. Consequently, organisations end up with AI solutions that are technically sound but unused, or useful but risky.

Nobody decides what happens with the time AI saves: When an employee saves two hours a week, those hours rarely become visible. Instead, more meetings, more emails and more requests absorb them. The organisation pays for the licence but never sees the return.

Organisations reward speed more than learning: Fast rollouts look impressive. At the event, for example, one company shared its experience of going all-in on AI and rebuilding its engineering team around AI-native ways of working. The change brought opportunities, but also resistance, rising costs and governance challenges they had not expected. One of their main lessons was that they had moved too fast.

 

 

 

My model: four building blocks that turn AI access into value

 

To move from access to value, organisations need four building blocks. Each one answers a question that licence numbers cannot.

 

Building block 1: Design trust into daily work

Trust grows when people can understand, question and influence what AI produces. Four questions from the event capture this well:

• Meaning: Do I understand what I see?
• Decision: Am I applying critical thinking?
• Control: Can I act on my judgement?
• Feedback: Does the system respond to my input?

Take a sales manager who receives an AI-generated forecast. First, she needs to know which data the forecast uses (meaning). Second, she needs to recognise when to challenge it, for example when a key customer has just changed its strategy (decision). In addition, she must be able to adjust the forecast without a complicated workaround (control). Finally, her corrections should visibly improve future forecasts (feedback). If employees cannot answer these four questions positively, even the best AI solution will struggle with adoption.

 

Building block 2: Look at every use case from three perspectives

A successful AI use case brings together three priorities:

• Business value: Does it solve a relevant problem and deliver measurable results?
• Rights and governance: Does it respect regulations, fundamental rights and the organisation’s responsibilities?
• People: Does it support employees, preserve their sense of purpose and help them do meaningful work?

You need all three. After all, nobody would call a technically impressive solution successful if it creates anxiety, lacks governance or fails to deliver business value.

 

Building block 3: Decide what to do with the time AI gives back

One question from the event stayed with me: what should organisations actually do with the time AI saves? Filling every saved hour with more tasks is only one option. In fact, leaders can choose between four:

• Do more: Improve service quality and spend more time on complex work and human interaction.
• Remove work (do less): Eliminate unnecessary activities and allow some of the efficiency gain to become free capacity.
• Shift capacity (do things differently): Move capacity to areas where it is needed and help employees build new skills.
• Improve (do things better): Invest time in learning, development and improving processes.

This choice belongs to leadership. Technology can create the time, but only leaders can decide how teams use it.

 

Building block 4: Set the pace on purpose

Starting small is good advice, but small steps only lead to progress when they are coordinated. AI transformation therefore needs experimentation, clear priorities and the ability to learn along the way. A good pace also allows you to notice resistance, rising costs and governance gaps early, while they are still easy to fix.

Change what you measure. Replace licence and training numbers with indicators that show value. For example, you can track:

• the time a key process takes from start to finish
• error rates or rework in AI-supported tasks
• the share of AI use in core workflows, beyond emails and summaries
• where teams have reallocated saved time
• how confident employees feel when working with AI

 

 

 

What corporate leaders can do now

 

• Remove licence counts as your main AI success measure and introduce three to five value indicators linked to business goals.
• Ask your teams the four trust questions (meaning, decision, control, feedback) and act on the weakest answer first.
• Review your current AI use cases from three perspectives: business value, rights and governance, and people.
• Decide with your managers what to do with saved time, and make that decision part of every new AI initiative.
• Link paid AI licences to defined use cases with an owner and measurable outcomes.
• Limit the number of initiatives to what your organisation can coordinate and learn from.
• Plan reviews every 90 days to adjust priorities, costs and governance as you go.

 

We spend a lot of time asking what AI can do. Yet the more important questions are what AI should improve, how we help people work confidently with it, and how we will know whether it creates value. Ultimately, those questions turn AI experiments into business results.