Is AI Delivering the Value You Expected?
AI has been accompanied by some significant promises: It will reduce costs, save time, release capacity, improve decisions and help us do more with less. Yet KPMG's latest Global AI Pulse found that only 26% of organizations have real-time visibility into the cost of running AI. The same research points to a wider challenge: translating growing adoption into measurable and sustainable business value.
That raises an obvious question - if an organization cannot clearly see what AI is costing, how confidently can it say what value it is creating?
Use does not prove value
Many organizations can show that AI activity is increasing. More people are using approved tools and running pilots. New features are frequently being added to everyday software. Employees are likely producing work faster, at least in some areas.
All of that may be useful but it still leaves the harder question unanswered - what has improved for the business?
An employee completing a task 30 minutes faster does not automatically create a financial return. A pilot producing an impressive result does not mean the organization is ready to scale it. High adoption does not necessarily indicate that performance has improved.
Time saved becomes valuable when the organization makes a deliberate choice about what happens to it.It might allow a team to serve more customers, improve the quality of its work, respond more quickly, reduce reliance on external support or focus on higher-value activities.
Without that next step, the saving may remain largely theoretical.
Start with the intended value
One reason value is difficult to measure is that organizations often begin with the technology. They introduce an AI tool, encourage adoption and then look for evidence that something useful has happened.
A better starting point is to define what you expect to improve. That might mean better quality or consistency in an output, more capacity for higher-value work, an improved customer or employee experience, reduced operational or compliance risk, or stronger revenue. Each of these needs a different measure.
A tool intended to reduce processing time should not be judged solely by the number of employees using it. An initiative designed to improve decision quality may not produce an immediate cash saving. An early experiment may be valuable because of what the organization learns, even if it is not yet ready to scale.
Ultimately, the measure should reflect the purpose.
What does AI really cost?
The visible cost may be a licence, a platform or an external partner. The fuller picture can include implementation, integration, training, experimentation, governance, output checking and the time people spend adjusting how work gets done.
Not every initiative needs a complex financial model. Leaders do, however, need enough information to compare the expected benefit with a realistic view of the investment.
A claim that AI has saved 500 hours means little without context. Whose time was saved? How was it calculated? Was the capacity used differently? Did quality remain consistent? Were new checking or correction tasks created elsewhere in the process?
Three questions for leaders
For any significant AI initiative, leadership teams should be able to answer three questions:
What did we expect, and what evidence shows it's happening? Was the aim to reduce cost, release capacity, improve quality, generate revenue, reduce risk or build capability? Clarity at the beginning makes it possible to recognise the evidence later, whether that's usage data pointing to real outcomes or the initiative still running on assumption.
What did it actually take? Not every initiative needs a complex financial model, but leaders need enough of the real picture - implementation, training, governance, workflow change - to compare against the benefit honestly.
What will we do with what we learn? The evidence should lead somewhere: scale it, redesign it, keep learning, or stop.
A question worth asking
Choose one important AI initiative in your organization and ask:
What value did we expect, what has actually changed, and what evidence would justify our next decision?
A clear answer will tell you that the organization has a useful basis for deciding what happens next. A vague one suggests that the value may still be assumed rather than demonstrated.
