Further random thoughts on AI in retail
No one doubts the technology can be transformative, but has it delivered for retail as expected yet?
(Source: Pexabay)
Do we have doubts yet? Here we are in 2025, three years into what some call the AI Hype Cycle, and discussion is now turning towards whether the billions of dollars poured into AI and its associated infrastructure constitutes a bubble on a par with the dotcom frenzy of the late 1990s.
Only this week, the Bank of England warned of the possibility of a “sharp market correction” should investor confidence in AI tech giants suddenly evaporate, which was hardly surprising given how much the US economy is currently being fuelled by investment in the technology.
Racing towards change…because we must
Worldwide, businesses of every stripe are transfixed by the AI debate: should they invest? And, if so, how much should they invest? AI evangelists promise transformational operational rewards for those that believe; entire companies are restructuring their organisations with an eye to maximising AI potential.
Upheaval is everywhere, and the discussion is increasingly becoming polarised, depending on the business. In the most extreme instances, employees are mandated to embrace AI or risk being ousted. In companies where committed true believers hold sway, literally nothing is deemed impossible. Headcount is being reduced in the steadfast conviction that AI can perform that role – even when actual proof of that assertion has yet to be demonstrated.
The retail industry has not been immune to AI mania, but for all the collaborations with major providers and expensive projects launched to date, the quantifiable impact thus far has been limited. A new survey from German consultancy Horvath puts AI’s positive impact on sales at 0.4%. And where respondents do see benefits from AI application are in personnel costs in areas like CRM, marketing and sales, with 8% gains anticipated on average.
Yet other studies indicate these savings may carry another kind of cost. Qualtrics’ 2026 Consumer Experience Trends Report found that one in five shoppers interacting with customer service AI were dissatisfied with the experience. A further 53% expressed concern about potential data mishandling while around half of respondents lamented the absence of human interaction.
Read more Retail Slop on this topic: Hype or hope? Random thoughts on AI in retail
Anyone that has studied the potential AI holds to radically overhaul numerous areas of the industry understands there are big wins to be had. What appears to be the pressing question right now is how soon these wins will manifest and whether the industry is positioned to capitalise on the technology’s promise.
Systematic failures?
This week, the UK Groceries Code Adjudicator flagged serious breaks in supermarket forecasting systems that meant suppliers were experiencing losses due to poor practices, some of which was connected to AI implementation across supplier portals. Inaccurate forecasting was a major problem, with “unclear, unreliable and unusable” data deemed a serious issue.
This illustrates the dilemma inherent in AI rollouts. Companies are accelerating AI projects in pursuit of quick wins that benefit the bottom line. Such initiatives may be urged by an especially-driven AI-evangelist CTO or particularly demanding management determined to get the jump on rivals in the space.
However, retailer IT systems are typically complex beasts, often with multiple architectures working in conjunction. Introducing a completely new system and expecting it to work straight out of the box is not as easy as it looks. And it’s not a simple or inexpensive process to persuade legacy systems to play nice with the new shiny interloper.
There may also be a drive to deliver results to justify investment to shareholders. Everyone in the industry acknowledges these business instincts, but everyone is also familiar with grandiose projects that flop due to a failure to look beyond initial proof of concept.
The retail industry has a long history of initiatives that are implemented at speed only to disappoint once in place. Consider self-checkout rollouts that just annoy shoppers; in-store café builds quickly ripped out as consumer habits change, or retailer ventures into alternative channels or new markets that fail to deliver.
You can be sure every one of these ideas was fully scoped and greenlit from the c-suite down, with hundreds of employees involved, only to come to naught because one crucial detail was overlooked in the rush to execute on what everyone agreed was an infallible vision. That detail may have been operational or simply a failure to understand what the customer actually wanted.
A hard and steep learning curve
And this may be the risk with AI. We know that failure rates in the space are high, as much as 95%, but then there is an additional issue that rarely gets addressed – AI is very much a new technology. Sure, we can be pedantic and point out that artificial intelligence has been developed over decades, and this true, but in real-world terms it is pretty much in its infancy.
It may simply be the case that genuine expertise in the field is currently limited, and what high-level adepts in mastering AI exist will more likely be working for far higher wages at the likes of OpenAI or Anthropic than plying their trade in the IT department of a mid-sized European or US retailer.
This is not to decry the ability of the legions of uber-coders hunched over their laptops at retailers worldwide; it just seems more likely many of these are often learning AI skills on the fly. Meanwhile, they are also ensuring the company’s legacy systems function as intended. AI expertise will scale fast in the coming years, but the question right now has to be does enough talent actually exist to facilitate expert execution at the required scale?
And so we wait. As things stand, we see endless retailer announcements touting a new “AI-powered” initiative, many of which, on closer inspection, appear to be simply machine-learning in fancy new clothes. The most disruptive retail impact of actual AI thus far has been its impact on online search, with the rise of ‘zero-click’ search overturning long-accepted rules on how ecommerce operates. This has thrown the SEO industry into turmoil as experienced digital marketers scramble to get their heads around wholly new concepts like Generative Engine Optimization.
AI’s potential is undoubted, and we are seeing results, primarily in back-end operations and areas like supply chain optimisation. But the notion that, one day, the technology will miraculously be able to predict shopper product choice and somehow materialise that product in their hand before they knew they want it (and I have heard some say this), seems far-fetched. Because machines and data can be managed for predictable outcomes, but humans cannot.
Perhaps a better strategy would be to concentrate on what the humans, with all their foibles, want and instruct the machines to better serve that, rather than the other way around. It’s just a thought.
If you enjoyed this content, be sure to subscribe free to Retail Slop on Substack for regular updates and insights on developments across the retail world.