Why AI will not fix broken customer operations

Most CX automation starts with volume. Find the largest queue, calculate the cost and point AI at it.

That looks rational. It is also how companies automate interactions that influence revenue, retention or a customer’s willingness to trust the brand. High volume does not mean low value.

Some interactions are predictable and should disappear into self-service. Others require judgement, negotiation or a human who can recognise what is really at stake. Treating them as equal is not transformation. It is indiscriminate cost cutting.

A customer journey map is no longer just a picture of touchpoints. In AI implementation it becomes the diagnostic: how work really moves, where customers wait, which handoffs create rework, where knowledge fails and which decisions require judgement.

Then assess the value of each interaction. Does it protect revenue? Could it prevent churn? Does the employee need to negotiate, interpret ambiguity or make a decision with consequences?

Those answers change the design. Low-value repetition is a strong automation candidate. High-value judgement may be better augmented: give the employee better knowledge, faster context and clearer options while keeping accountability human.

The point is not to protect human work for sentimental reasons. It is to redesign the whole journey and use AI where it improves the outcome rather than merely removing effort.

Once the interaction is understood, look underneath it. If the process has conflicting steps, AI scales the conflict. If the knowledge base is outdated, AI distributes outdated answers with more confidence and speed.

Process and knowledge are usually the first things broken. Fix ownership. Remove unnecessary handovers. Decide which source is trusted. Make exceptions explicit. Only then can AI become part of a dependable operating model.

A fluent answer is not the same as a correct resolution.

Real-time translation changed the footprint available to Samsung Nordics and Transcom. But translation alone did not create the result. The operation was ready: the workflows, knowledge, governance and quality discipline could support a different delivery model.

That combination made the footprint redesignable, contributed to a 15-point NPS improvement and was recognised with an ECCCSA Silver. It also supported a lower-cost footprint without lowering the ambition for customer experience.

The useful lesson is not “buy translation”. It is that AI creates leverage when the surrounding operation can absorb it.