User testing a new customer-contact tool for a Dutch government organization

The Tangle

A new customer-contact approach was being developed to help employees find the right information more intuitively.

Rather than searching primarily through large amounts of information, the intended logic was to start with the citizen’s actual question and use that context to guide the employee towards the relevant answer.

The challenge was that real customer questions contain nuance. Small differences in timing, tax situation or terminology can completely change which information is correct.

The Disentangler

As a user of the customer-contact environment, I participated in testing the new approach from the perspective of the employee who would actually need to rely on it during a customer interaction.

My focus was not only whether the system technically returned an answer, but whether it correctly understood what the customer was really asking.

That distinction matters: an answer can look plausible while still being wrong for the specific situation.

The Disentangling

During testing, I worked through realistic questions and examined how the system interpreted them.

That exposed cases where contextual distinctions were not yet being recognised reliably. For example, similar terminology could lead the system towards information about a final assessment when the actual question concerned a preliminary assessment for the current year.

Those kinds of cases are valuable because they reveal where system logic, language interpretation and real operational practice do not yet align.

The feedback from testing can then be used to improve question recognition, decision logic and the reliability of the information presented to employees.

The Disentangled Result

The value of testing was not simply finding individual errors.

It was helping identify where the intended customer journey broke down when confronted with real-world language and context.

That is the point where digital transformation becomes practical: not when a new system technically works, but when employees can trust it during actual customer conversations.

A system does not need to understand every word. It needs to understand the right question.