RTS 2024 and the Enterprise–Mid-Market Divide

Twelve and a half thousand retail professionals. Four hundred exhibitors. Olympia London, which has the slightly overwhelming quality of a building designed for Victorian industry exhibitions that has been hosting technology events without quite acknowledging the mismatch ever since. RTS 2024 was, as ever, a big, loud, exhausting, and occasionally genuinely interesting event.

The dominant theme was predictable. AI and security emerged as the twin pillars of the conference, unsurprising given where the industry had landed by early 2024, but the way the AI conversation played out was more complicated than any single theme captures.

The Main Stage vs. the Floor

The tension that ran through RTS 2024 was between the case studies presented on the main stage (from Primark, Boots, Morrisons, Harrods, Adidas, Nobody's Child and others) and the reality of what most of the 12,500 attendees could actually deploy. The enterprise retailers presenting had substantial data infrastructure, dedicated AI/ML teams, and the organisational capacity to run multi-year programmes that produce impressive, polished case studies. The mid-market buyers walking the exhibition floor were trying to figure out what they could do with limited technical resource and a competitive need to do something.

This is not a new tension in retail technology. It existed through the first wave of ecommerce platforms in the 2000s, and before that in the supply chain technology debates of the 1990s. But it feels more acute now because the capability gap between what AI can do for a well-resourced enterprise retailer and what is accessible to a retailer with a small digital team is, if anything, wider than in previous technology cycles.

The Shopify democratisation argument, that AI tools built into platforms like Shopify and Klaviyo are making enterprise-grade capability available to independent retailers, was audible on the exhibition floor. But the practical reality is that deploying even well-packaged AI tools effectively requires data quality, process discipline, and some level of technical capability that is not evenly distributed. Platform AI lowers the floor; it does not eliminate it.

What Was Actually on the Show Floor

The exhibitions were a useful indicator of where real commercial deployment is happening versus where speculative development is being demoed. A few consistent patterns stood out.

Workforce management AI was everywhere, and the conversations around it were notably more grounded than the customer-facing AI discussions. Orquest, one of the exhibitors, was discussing AI-driven scheduling optimisation that directly addresses the labour cost challenge facing UK retail: rising minimum wages, variable footfall, the complexity of managing part-time and zero-hours contracts efficiently. The ROI case for this is clear, the data requirements are manageable, and it is in production at scale. Less glamorous than conversational shopping AI, considerably more widely deployed.

Computer vision for inventory and loss prevention was a consistent presence. Not the fully autonomous checkout systems that generate NRF-level press coverage, but the more bounded application of cameras and AI to accurately detect what is on shelves, identify mis-picks, and improve inventory accuracy. This is the version of physical retail AI that is actually getting deployed in UK stores, rather than the grab-and-go vision that tends to dominate the coverage.

Customer data platforms and personalisation tooling were the most contested space on the floor. Lots of vendors, lots of claims, and considerable variance in sophistication. The more credible conversations were about data integration and data quality as the primary challenge, with AI personalisation as the downstream application once those problems are solved. The less credible ones were promising AI personalisation without much discussion of what data it would run on.

The Honest Conversation

One session I found useful was a retailer-led panel on AI implementation explicitly structured around what had not worked as well as what had. This is rare at commercial events, where the incentive structure pushes toward success stories, and the honesty in the room when the format invited it was striking.

The recurring themes: data quality is the gating constraint for most AI applications, not model sophistication. The integration work required to connect AI tools to existing commerce infrastructure is consistently underestimated in project plans. Organisational change management (getting merchandising, buying, and digital teams to trust and act on AI recommendations) is often harder than the technical implementation.

None of this is surprising if you have been around long enough. But it is useful to have it confirmed by practitioners rather than inferred from vendor materials.

RTS 2024 was a useful read on where the UK retail technology industry actually stands: more advanced than it was three years ago, more cautious than the conference keynotes imply, and genuinely uncertain about which AI bets will pay off at what scale.