Amazon Connect is taking aim at one of the contact center industry’s longest-running obsessions: deflection. In a recent AWS Pulse discussion, AWS leaders argued the goal is no longer to keep customers away from humans. It is to build stronger relationships when customers do reach out.
The message lands at a tense moment for CX teams. Many leaders still run contact centers like cost centers. But customers now expect faster resolution across voice and digital. And they punish brands that hide behind rigid automation.
Tony Gooch, Senior Manager of Product Management at AWS emphasized how contact centers win loyalty when they show up for customers during high-stakes moments framing the shift as a practical shift:
"Getting the customer engagement right on the most critical transactions is really one of the most foundational things to enduring long-term customer longevity."
Why “Deflection” Is Losing Its Grip
For years, deflection was treated as the cleanest measure of efficiency. Fewer calls meant lower agent costs. But Gooch argued that approach can miss the real point of service, especially when a customer’s issue is urgent or emotionally charged.
He shared an example of a young person who rarely made phone calls, but immediately called support when a new PlayStation did not work. The point was simple. Customers may avoid voice most of the time. But when the moment matters, they still want a real outcome.
That is why Amazon Connect is pushing a different framing. Relationships are the target. Deflection is a tool at best, not the outcome.
“AI That Helps,” Not AI That Replaces
The episode also highlighted how AWS sees AI changing frontline work. The focus was not just on automating conversations. It was on lifting agent performance and improving customer outcomes during live interactions.
Gooch described how enterprises often have a wide performance spread across agents. In his experience, a “perfect agent” can sit far above the average. He argued that agent assist can help close that gap at scale by putting knowledge, guidance, and suggested responses into the agent workflow in real time.