AI Alone Won't Transform Customer Service, EY Report Finds

A new EY report finds that deploying artificial intelligence tools on their own is not enough to transform customer service — organizations also need to redesign how their operations actually work. The report argues that bolting AI onto existing processes, without rethinking workflows, roles and decision-making, limits the impact companies can expect from their technology investments.

EY's analysis points to a gap between AI adoption and AI-driven outcomes: many companies have rolled out chatbots, virtual agents or automated routing, but few have restructured the underlying operating model that determines how service is actually delivered. The report frames this as the difference between layering technology onto legacy processes and genuinely redesigning them around what AI makes possible.

This is a timely corrective for leaders who have treated AI adoption as a checklist item rather than a structural change. The finding suggests that return on AI investment in customer service depends less on which model or tool a company buys and more on whether it is willing to re-architect processes, staffing models and escalation paths around that technology.

For digital transformation leaders, the implication is that AI strategy and operating-model redesign cannot be separate workstreams. Without the latter, AI risks becoming a thin layer of automation sitting on top of the same friction-filled processes customers already experience — faster, perhaps, but not fundamentally better.

This report confirms something René sees repeatedly in the field: technology adoption is treated as the finish line, when it should be the starting gun.

Most organizations buy AI the way they'd buy a faster printer — expecting speed without expecting change. But service design doesn't work that way: if the underlying process still routes a frustrated customer through five handoffs, AI just gets them to disappointment faster. The real unlock isn't the model, it's the courage to kill the workflow the model was bolted onto. Operators who are serious about this should pick one painful, high-volume journey, strip it down to the decisions and handoffs that actually matter, and only then decide where AI belongs in it — not the other way around.