Table of Contents
Most retailers want to understand how AI and customer experience in retail fit together. They think they know how their customers feel about AI. Data from UK consumers suggest that confidence may be misplaced.
YouGov’s 2025 survey of over a thousand UK consumers found that two-thirds are comfortable with AI comparing prices, and nearly half accept personalised deal suggestions from AI. But trust drops sharply when AI moves from informing customers to acting on their behalf. Only one in three trusts AI to handle a complaint. One in ten trusts it to place an order for them.
Mintel’s 2026 consumer AI report finds the same pattern. Customers draw a consistent line between AI as a tool and AI as a decision-maker. A significant amount of current experiences using AI either sit on the wrong side of that fine line or are very close to it.
Why AI and Customer Experience in Retail Are Diverging
The decision to automate customer-facing interactions is rarely made lightly. ContactBabel’s UK CX Decision-Makers’ Guide 2026 shows that the average inbound customer service call now costs retailers around £6.25, up nearly 50% over five years. At that cost, automated channels aren’t an indulgence; for many retailers, they’re a financial necessity.
But the economics of deployment don’t determine the quality of the experience. Research from ArvatoConnect (conducted in financial services rather than retail) found that nearly three-quarters of vulnerable customers had felt like giving up entirely when using automated service channels. Only two per cent said those channels consistently gave them what they needed.
The systems were working exactly as designed. The problem is that for customers whose situations didn’t quite fit the standard options, there was no easy way to reach a human who could exercise their judgment and help sort it out. That’s a design failure. The customer’s experience obviously wasn’t the primary aim when the system was being built; cost reduction was.
This is exactly the kind of gap that’s opening up between AI and customer experience in retail right now.
The Privacy-Relevance Tension Retailers Are Getting Wrong
AI and customer experience in retail often clash when it comes to personalisation. Retailers want to use customer data to make experiences feel more relevant, but customers worry about privacy and how their data is used.
So the business case for personalisation is often too simple: it assumes “people want relevance, so they’ll accept data use,” when in reality customers are more cautious, and the trade-off is often tighter than retailers assume.
Braze’s 2026 research found that fifty-eight per cent of UK consumers want experiences that feel personally relevant. Nearly half say data protection is the main thing they’d need before trusting an automated system. A third won’t share personal data with AI at all.
Retailers tend to resolve that tension by assuming the desire for relevance wins. The research suggests it’s far more balanced than that. When personalisation tips from helpful to intrusive, most customers don’t complain; they disengage.
They give a polite score and don’t return. It registers as broadly positive in the data, but something has already changed.
This is the same gap explored in Episode 20, which examines whether surveys are actually working for VoC.
There’s a difference between what customer listening systems capture and what customers actually experience. The customer’s absence is invisible until it’s flagged in footfall or frequency data, by which time, it’s too late.
What AI Insight Tools Are Doing to Customer Curiosity
The subtler risk sits inside retail CX teams themselves, not in customer-facing channels. This is another place where AI and customer experience in retail are pulling in slightly different directions.
Most large UK retailers now run AI-assisted insight tools alongside their VoC programmes. Platforms like Medallia and Qualtrics surface weekly summaries that include themes from verbatims, complaint spikes, and sentiment trends, without requiring anyone to read individual comments. For stretched teams, this is a genuine operational improvement.
But it changes what gets prioritised. When the dashboard produces a comprehensive-looking weekly summary, the question being asked isn’t when are we next talking to customers directly? It becomes what would that tell us that the dashboard isn’t already showing? The qualitative research programme keeps getting pushed back, because it’s harder to justify the extra investment when the AI output already looks thorough.
What gets lost is the unexpected finding. The verbatim that didn’t fit any category. The store visit where a customer did something the journey map said they wouldn’t. AI insight tools process what’s already coming in. They can’t go looking for what isn’t.
If you haven’t listened already, Episode 29 explores how insight gets generated and why it matters as much as what you do with it once you have it.”
Thinking Like Your Customer Is a Present-Tense Discipline
Bill Stinnett’s Think Like Your Customer was published in 2006, but its central argument applies directly to how AI and customer experience in retail are evolving right now. Stinnett’s point is that most business professionals default to their own frame of reference; their processes, their logic, their internal knowledge. The deliberate act of setting that down and asking, “What does this look like from where the customer is standing?” doesn’t happen naturally. And the more expert someone becomes in their own business, the harder that act becomes.
AI has the same structural problem. It builds its picture of customers from past behaviour, i.e. purchases, clicks, returns. That picture can be detailed, well-modelled and entirely out of date. The customer who bought running gear last spring may have since had an injury. Another customer may have changed jobs, and their income has changed. Another customer may simply have moved on. The algorithm doesn’t know context. It only knows what happened.
The specific risk is that AI presents this historical picture with confidence. The output looks current, clean and authoritative. It gets used as a description of who customers are rather than a hypothesis about who they were. AI can tell you what happened and predict what might happen next. It cannot tell you who your customer is right now. As Stinnett frames it, thinking like your customer is a present-tense act. That’s still something only a human can do.
It’s worth being clear that AI does improve customer experience in specific, well-evidenced contexts.
ASOS’s AI stylist works because the product data infrastructure behind it is mature. Robert Dyas replaced expensive on-location photography with AI-generated imagery for home and garden products, resulting in significant cost savings and improved functionality. Both work because the AI is doing what it’s built for: pattern recognition, optimisation, scale. Neither requires AI to understand what a human (customer) means.
Three Things Worth Checking in Your Own Organisation
If you’re thinking about AI and customer experience in retail in your own organisation, here are three things worth checking.
First, when did your team last do research that wasn’t a reaction to something the data flagged? If you can’t remember, or if the honest answer is “we only investigate once the AI tells us something’s wrong,” that’s worth noticing. It means the AI is now deciding what gets looked into, not your team.
Second, when AI insight gets presented to the people making decisions, does anyone ask what it might be missing? Things like: what context isn’t in this data? Is there a reason for this pattern that the AI wouldn’t know about? If insight goes straight from the dashboard to a decision without anyone asking those questions, that step is being skipped.
And third, how are AI-generated customer profiles being used? These profiles are built from past behaviour, so they’re a reasonable starting point. The risk is treating them as a current, accurate picture of your customers rather than a guess worth checking. That’s the same mistake as assuming your own expertise is automatically up to date.
Key takeaways
- UK customers have drawn a clear line: AI as a tool, yes. AI making decisions for them, no. A lot of what retailers are currently doing is either very close to, or crosses that line.
- Automated service channels work for straightforward queries but fail at the edges (like unusual problems, vulnerable customers, or situations the system wasn’t built for). That’s a design choice, not a technology limit.
- AI insight tools don’t replace talking to customers directly; they just make it easier to skip that step. And the things you’d only find by looking (the comment that doesn’t fit a category, the thing a customer did that surprised you), stop getting found.
- A customer profile built from past behaviour is a guess about who that customer used to be. Treating it as a description of who they are now is where things go wrong.
- Thinking like your customer means staying curious about who they are right now. AI can only ever tell you who they were.
