Glad, Mad or Sad? A Human Test for AI-Shaped Customer Journeys

Most customer-journey maps are beautifully rational. Boxes, arrows, channels and hand-offs suggest a customer moves through the organisation as planned. Real people do not experience journeys that way. They remember moments of relief, irritation, uncertainty and care. They also act on those emotions: they continue, abandon, complain, comply, recommend or try again.

Behavioural science helps explain why. Research on the peak-end effect indicates that people do not simply average every moment of an experience; intense moments and the ending can shape what remains in memory. Research on negativity bias further suggests that a poor experience can carry more weight than an equally positive one. A journey may therefore meet its service-level targets and still leave a customer feeling unheard.

AI makes this gap more important. It can reduce waiting, retrieve information and personalise an interaction. It can also route the customer into the wrong queue, present a confident but unsuitable answer, restrict an exception or ask someone to repeat what the organisation already knows. The technology does not merely improve or damage a touchpoint. It changes what people notice, believe and do next.

Two journeys, one experience

There are always at least two humans in a service journey: the external customer seeking an outcome and the internal employee trying to produce it. Their journeys are linked.

When AI handles routine queries, employees may inherit a higher concentration of complex, emotionally charged or unresolved cases. The customer arrives frustrated. The employee opens several systems, interprets an AI recommendation, follows a policy, manages time pressure and absorbs the customer's emotion. If the employee cannot question the recommendation or explain the reasoning, both sides lose agency. Behaviour can swing between over-reliance on automated advice and rejecting it after a visible error. Neither response is calibrated judgement.

That is why employee experience is not an internal HR issue separate from customer experience. It is part of the delivery mechanism. Cognitive load, unclear decision rights and low psychological safety inside the organisation can become repetition, delay and defensiveness outside it.

A Glad-Mad-Sad retrospective

I would add a simple retrospective to the conventional journey map. At each important stage, ask the customer and the employee three questions.

Glad: What created relief, confidence, ease or a sense of control? Perhaps AI surfaced the right history, prevented repetition or gave the employee time to listen.

Mad: What created frustration, unfairness or resistance? Look for forced channel switching, unexplained decisions, rigid scripts, inaccessible language and AI outputs that could not be challenged.

Sad: What created disappointment, helplessness or loss of trust? This is often quieter than anger. It may appear when the customer gives up, the employee stops escalating a known problem, or both accept that "the system will not allow it".

The value is not the emotion label alone. The next question matters more: What behaviour followed? Did the customer abandon, repeat contact, escalate, withhold information or move to another provider? Did the employee comply without checking, override informally, create a workaround or disengage?

This turns emotion into operational evidence. It also surfaces procedural fairness. People may not always receive the outcome they want, but voice, explanation and a credible route to review can change how the process is experienced. AI should therefore not only optimise the answer. It should preserve the conditions under which the answer can be understood and questioned.

From a touchpoint map to a behavioural system

A useful review should connect every Glad-Mad-Sad moment to the decision that preceded it. Was the decision made by a person, a rule, an AI model or some combination? Who was accountable? Who could challenge it? What did the organisation learn from the exception?

This is also where middle managers matter. They see patterns across complaints, transfers, overrides, workarounds and coaching conversations. If they are measured only on speed, containment and compliance, valuable behavioural signals may be suppressed as inefficiency. If they are equipped to interpret them, recurring emotions become clues to broken knowledge, policy, process or technology.

A genuinely 360-degree assessment therefore looks beyond the visible customer interface. It examines the relationship between strategy, people, processes, technology, data, customer interactions, quality, risk and the environment in which staff work. The purpose is not to make every customer glad. That would be unrealistic. It is to understand why people become glad, mad or sad, what they do next, and whether the system can learn.

Before buying the next AI tool, run this retrospective across both journeys. The most important insight may not be where AI saves a minute. It may be where a customer loses trust, an employee loses discretion, or the organisation loses the signal that something needs to change.