When AI Forces CX Leaders To Confront Hidden Process Failures, Everyone Wins

AI has a way of cutting straight through customer service theater. When organizations deploy it, the technology quickly removes the human judgment and improvisation that once kept imperfect systems running. What emerges is an unfiltered view of how work actually gets done, where rules break down, and which processes only function because people compensate for them. In that role, AI becomes a forcing function, bringing long-hidden operational debt into the open and demanding clarity where there was once flexibility.

Aurelia Pollet is the Vice President of Customer Experience at CarParts.com, bringing over 20 years of experience across eCommerce, retail, and technology. Her perspective comes from a uniquely diverse career, including leadership roles at the pinnacle of luxury service with Louis Vuitton and the high-volume worlds of Quest Nutrition and Exemplis LLC. She sees technology as a stress test for CX, showing where systems are resilient and where they depend on human heroics.

"For years, our call center agents absorbed the problems in our systems and worked around them. When you put AI on top of it, you can’t do that anymore," says Pollet. The proof is in a classic business headache: product returns.

• Headlight headaches: Returns, she says, rarely start at the return desk. They begin earlier, when customers are forced to make technical decisions without the knowledge to do so. "If you don’t know there are differences between headlights, you’re not going to look for the right things because you don’t know what you’re looking for," Pollet explains, pointing to how even small product variations can derail a purchase. Rather than optimizing returns after the fact, she argues the real fix is upstream: an AI sales assistant that behaves like an expert, slows the moment down, asks the right questions, and helps customers get the decision right the first time.

• Off the rails: AI doesn’t improvise the way people do. When a process is unclear, it breaks. Pollet points to a moment when that became obvious: "An AI agent told a customer to drop a return off at one of our warehouses, and the warehouse called asking why we were telling people to show up there," she recalls. "That’s when you see how this can get out of hand. It has a mind of its own, and you need to put guardrails in place." The example is small, but the signal is clear. When AI is allowed to operate inside loose or poorly defined processes, it exposes the gap between what an organization intends and what its systems are actually equipped to handle.

• Delight not required: As AI pushes CX leaders upstream, Pollet argues it also forces a reset on what success actually looks like. "There’s this industry obsession with having to delight the customer, and if I hear that word one more time, I’m going to scream," she says. "In luxury, delight is the goal. But when you’re buying a headlight because you had an accident, you don’t need delight. You just want the part to come as fast as possible." In most transactions, she believes efficiency is the real form of respect. Get the right product, deliver what was promised, make returns easy, and save human empathy for the moments that truly require it.

The takeaway is that AI governance can’t live on an IT or legal checklist. It belongs with customer experience. When an AI speaks to a customer, its tone, accuracy, and decisions are the brand in action. Making that real requires a clear ownership model, with CX setting the behavioral rules and working in close partnership with data science and technology to define boundaries, escalation paths, and risk tolerances that hold up in the real world.

• Wipe the slate clean: For established companies weighed down by legacy systems, Pollet sees AI as a rare reset button. "With AI, you don’t have to think with constraints," she says. "You can challenge your assumptions and ask whether a process should even exist, or whether it could be completely different." Rather than patching broken workflows or optimizing around old limitations, she argues AI creates space to redesign from a blank page, replacing incremental fixes with fundamentally new ways of operating.

Pollet saves her strongest argument for leaders still hesitating. The real risk, she argues, isn’t that an AI initiative might stumble. It’s missing the discipline the effort forces on the organization. "Let’s pretend the chatbot doesn’t work and we have to go back to full human," she says. "It’s still worth it because we went through every process, one by one. We killed things, we improved things, we changed things. That exercise alone was worth it." Even if the technology evolves or fails, the work of preparing for it leaves behind something far more durable: a clearer, stronger business built on intent rather than improvisation.