The fundamental error in recent AI customer service deployments has been treating queuing as an engineering problem rather than a psychological one. Klarna's announcement that its OpenAI assistant replaced 700 agents, Air Canada's chatbot liability case, and similar high-profile failures all share a common architecture: technology designed to answer faster and cheaper, not to address what customers actually experience when waiting. David Maister's 1985 framework—satisfaction equals perception minus expectation—remains the most predictive model for service outcomes, yet it is routinely ignored by implementations that optimise for speed. The Houston airport baggage claim case illustrates the principle perfectly: moving gates further away, so passengers walked six minutes instead of one before waiting two minutes instead of seven, eliminated complaints despite identical total time. What matters is not the duration of the queue but whether customers feel acknowledged, treated fairly, and certain about their position in the system. The psychology of waiting has not changed in four decades; what has changed is the complexity of the environment in which it operates.
Omnichannel support has multiplied the psychological burden by fragmenting the queue across channels and interfaces. A customer's journey now spans apps, chatbots, asynchronous messaging, callbacks, and human agents—each transition resetting their sense of progress and compounding uncertainty about whether the system knows they exist. This creates a critical operational question for CX teams: how do you operationalise Maister's principles at scale when customers are distributed across five different waiting states simultaneously? The answer emerging from successful deployments is not broader generative AI but tighter, more deterministic systems paired with real-time analytics. Bank of America's Erica—a narrowly scoped, honest-about-its-limits system handling three billion interactions—and Octopus Energy's AI drafting tool, which improved satisfaction from 65 to 85 percent by augmenting rather than replacing agents, both succeed because they do the psychology before the maths. Modern customer analytics platforms now operationalise this insight through journey analytics, behavioural modelling, and predictive abandonment detection—identifying the precise moment a customer decides they are being ignored or treated unfairly, then intervening before abandonment occurs.
The strategic implication for CX leaders is that the 33 percent failure rate Forrester predicts for new AI rollouts in 2026 will not be a technology problem but a design philosophy problem. With 85 percent of consumers still preferring humans for service, the question is not whether AI can replace agents—Gartner's prediction of 80 percent autonomous resolution by 2029 suggests it can—but whether your implementation honours fairness, reduces uncertainty, and demonstrates that someone is paying attention. Teams already running orchestration platforms like Zendesk or Salesforce have the infrastructure to detect these psychological inflection points; the gap is whether they are using it to predict and prevent abandonment or simply to route faster. The organisations that will calm queuing anxiety are those whose designers understand that a callback offer at minute six, routed to the right agent after two transfers, with a clear explanation of why the wait exists, will generate more loyalty than a chatbot that answers in seconds but leaves customers feeling unheard.
The Curious Psychology of Queues and the AI Quietly Trying to Calm Us Down CX Today
The Curious Psychology of Queues and the AI Quietly Trying to Calm Us Down CX Today