AI

In retail, automation isn’t the problem. Untested customer journeys are.

In retail, trust is built during the full customer journey, not during just the AI interaction. That’s why monitoring needs to become as central to contact center operations as the AI sitting on top of it.

Adobe Stock

July 20, 2026 by Mark Rohan — Co-founder and COO, Klearcom

Summarize with

AI proposes to vastly change industries by automating away grunt work, and businesses worldwide are buying into that promise. But few industries are as exposed to the masses as retail is. That is why conversations in the sector have been dominated by concerns about data privacy and AI ethics.

Forty-seven percent of consumers are ready to abandon brands over poor data retention and usage policies, and some even think AI actually worsens customer experience, according to a survey by Sogolytics.

The industry has responded with governance frameworks, transparency mandates, and data ethics guidelines. These are the right priorities, but the problem is that they address the wrong layer. When a customer gets routed to the wrong team or gets stuck in a loop they can't escape, no amount of algorithmic transparency can fix their experience. The failure is in the journey, and pushing more automation into broken flows only makes it worse.

Any AI tool deployed on top of your call infrastructure is only adding another layer of complexity to the system. Businesses must realize that with AI, their infrastructure now presents an entirely new set of challenges they can't afford to ignore.

AI may improve your system, but is it improving your customer's experience?

While contact center teams have always had visibility into how calls fail, AI decouples system performance from customer experience. An AI-powered contact center can report clean performance metrics while prompt changes or routing updates quietly degrade what customers reach. And that degradation will not register in the places teams are trained to look.

Standard QA wasn't designed for systems that change between tests. An AI-driven call flow can pass every test on Monday and fail customers by Friday — through no visible fault in the system. Old QA processes built around fixed, predictable customer journeys have no reliable way to catch that drift.

The most telling examples of systems silently failing are calls that connect but get routed incorrectly. Every internal metric will look okay because the call went through, the system did what it should, and the journey was completed.

But the customer who called about their undelivered order, dealt with an uncomprehending AI chatbot, and was bounced between departments before they got a resolution? They leave with an impression of a brand that can't seem to figure out what went wrong.

Pre-launch testing only proves the system worked once

A successful launch test is a snapshot. It confirms that a customer was helped successfully at a specific point in time, under the conditions that existed when the test ran. It says nothing about what happens six weeks later when an

AI tool's prompt is updated, a carrier reroutes traffic, or speech recognition struggles with a particular regional accent.

These variables don't come up during controlled tests. They happen when real customers call, and you can't catch a problem if you can't see it.

Retail brands need to test journeys the way customers experience them: from the first dial tone through IVR prompts, routing decisions, hold times, and final resolution. A connected call is not a completed journey, and system checks that stop at connection miss everything that matters after it. With AI in the picture, that bar rises further: tests need to verify not just whether a customer gets through, but whether they reach the right destination, with the right solution, within a reasonable time, and those checks need to run more frequently than the industry is currently used to.

Silent failures have loud consequences

Enterprise contact center outages can exceed $540,000 per hour, but operational costs can be mitigated. When AI fails a customer, the damage that ensues isn't always measurable in downtime: It shows up in reputational damage that lasts.

When UK parcel delivery company DPD's chatbot failed to help a customer locate a missing parcel, the customer asked it to write a poem criticizing the company. It complied. He then asked it to swear. It did that, too. The exchange went viral within hours. The chatbot had been updated in a way that stripped out its usual guardrails, and nobody caught the alignment drift until the damage was already public.

Klarna's experience runs deeper. The Swedish buy-now-pay-later company deployed AI that it claimed was handling the equivalent workload of 700 customer service agents, and CEO Sebastian Siemiatkowski later admitted the transition had negatively affected service and product quality. The company has since resumed hiring human staff.

The pattern reflects what happens when AI deployment outpaces the monitoring infrastructure around it: A model that performs well inside a broken or unmonitored call flow will still produce a broken experience. And the customer will blame the brand regardless.

What actually protects the customer journey

Retail brands are deploying AI into contact centers at a time when both the customer's tolerance for friction and the cost of switching to a competitor have reached rock bottom.

AI systems work well, and in some cases, are already offering up major returns — Airbnb said recently its AI customer service bot handles 40% of issues on its own. The problem is that internal performance and customer experience can't be measured with the same metrics, and AI investment tends to only widen that gap.

In retail, trust is built during the full customer journey, not during just the AI interaction. That's why monitoring needs to become as central to contact center operations as the AI sitting on top of it. In retail, the winners will not be the brands that automate the fastest. They will be the ones that can prove every customer journey still works.

About Mark Rohan

Mark Rohan is Co-Founder and COO of Klearcom, an AI-driven telecom testing platform headquartered in Waterford, Ireland with offices in India and the USA. He helps lead the company’s operations, customer engagement and global network strategy, supporting enterprise clients across 100+ countries in improving IVR, toll-free and voicebot performance.

Connect with Mark: