AI Service Failures: 82% of Shoppers Expect Brands to Make Good

What happened

According to a holiday report from Forethought published via Business Wire, 82% of shoppers expect brands to make good when AI-powered service fails to deliver what was promised. The headline frames the finding around broken AI promises, placing the burden of recovery on the brand rather than on the technology.

The release positions the research as a seasonal snapshot of consumer attitudes during the holiday shopping period, a time when retailers lean heavily on automated support, recommendations and self-service. The coverage available to us focuses on the headline finding, so the wider methodology, sample and specific failure types are not reflected here.

Why it matters

The finding shifts the question for leaders from whether AI can answer customers to what happens when it gets things wrong. Shoppers appear to treat an AI commitment as a commitment the brand has made, which means automated assistants now carry the same accountability as a human agent who misspeaks. For organisations redesigning service around AI, the operating model has to cover recovery as well as resolution: clear routes to a person, honest handover, and a defined way to put the error right.

For digital transformation programmes, this makes recovery a design requirement rather than an exception handler. Automation roadmaps that measure containment or cost per contact without measuring what happens after a failed AI interaction may be optimising the wrong outcome.

The Renascence take

The headline number is less interesting than the expectation behind it. Customers are not asking AI to be flawless; they are asking the brand to behave as though it owns the outcome, which is a basic fairness principle applied to software.

Most organisations will read an 82% figure as a reason to improve their chatbots. The more useful reading is that customers judge a brand by its repair behaviour, not its automation rate. Service designers should map the failure path first: how quickly a customer can reach a human, what that person is empowered to offer, and whether the brand admits the error without making the customer argue for a remedy. If your AI cannot make a broken promise right in one step, you have not finished designing it.