When AI Works, How Do You Know It is Creating Value?
Mon 24 Aug 2026
An AI system can be accurate, reliable, and widely used without answering the question that matters most to the organisation deploying it: what has improved because it is there?
That distinction runs through this week’s coverage in our wider Things That Matter editorial theme from the UK and German markets. Both perspectives move the conversation beyond whether AI works technically, but approach the value question from different directions.
In Germany, the experience of retailer Breuninger shows what happens when an organisation has to identify where AI creates enough value to warrant continued attention. The UK perspective starts one step further back, asking whether the measures organisations use to judge AI are capable of demonstrating that value in the first place.
Together, they expose a measurement problem that becomes more important as AI moves deeper into everyday operations: the easiest things to measure are not always the things that determine whether a project is succeeding.
AI Value Has to Be Found in Practice
The German perspective grounds the discussion in Breuninger’s experience and the lessons it has learned from putting AI to work.
Its experience points towards a more selective phase of enterprise AI adoption. Once organisations move beyond experimentation, the most important measure is whether that activity produces enough value to matter, not how much AI activity exists.
That makes measurement part of the decision-making process around AI itself. Organisations need evidence that helps distinguish use cases worth developing further from those where the return does not justify the effort involved.
Breuninger provides the practical perspective on that challenge. Its lessons show why value has to be understood in the context of the organisation using the technology rather than assumed from the capability of AI itself.
For enterprises with multiple potential applications competing for investment and attention, that differentiation becomes important. Success needs to provide enough evidence to inform what happens next.
The Dashboard Can Stop Too Early
The UK perspective examines the same problem from the measurement side.
Its starting point is that familiar AI metrics remain useful. Accuracy, adoption, speed, output, and cost can tell an organisation whether a system is functioning, being used and operating efficiently.
The difficulty comes when those measures are expected to demonstrate business value on their own.
The London analysis has argued that organisations need to follow what happens after the AI produces an output. A faster response only becomes meaningful when the wider process improves. Higher adoption matters when it changes how useful work gets done. Automation has greater significance when tasks are completed successfully rather than simply passed through an AI system.
That pushes measurement away from the model alone and towards the workflow around it.
It also makes baselines important. Without understanding performance before AI was introduced, organisations can observe what a system is doing without establishing what changed because of it.
AI Measurement Determines the Next Steps
The two perspectives meet at the point where measurement has to support a decision.
Germany brings the practical enterprise question: where is AI producing value worth pursuing?
The UK brings the measurement question: what evidence would allow an organisation to demonstrate that value credibly?
Neither can be answered by technical performance alone.
This changes the role of AI measurement. Metrics are no longer there solely to demonstrate that a deployment works. They also need to help leaders decide whether a use case should be scaled, redesigned, or stopped.
That requires different evidence at different stages. Technical teams still need to know whether systems are accurate, available, and secure. Operational teams need to see what is happening to workflows and human intervention. Business leaders need evidence that the resulting change matters to the organisation.
What becomes the challenge is connecting those views without allowing the most visible metric to become the definition of success.
Across this week’s German and UK perspectives, the common signal is that measuring AI becomes more demanding once the technology works. At that point, the question moves from whether the system can perform to whether its performance has changed something worth measuring.
