Two days at ExCeL London. Hundreds of conversations across retail tech, AI and transformation. Here's what we're still thinking about.
The Retail Technology Show always delivers a useful temperature check on where the industry actually is, not where the vendor decks say it should be.
This year, the conversations that kept coming up were less about what technology can do, and more about whether organisations are genuinely set up to use it.
I came away with ten things that stuck...
One of the sharpest moments of the event came courtesy of Steve Borg at White Stuff, who reframed transformation not as a final destination but as an ongoing organisational muscle to build. We've all heard the journey metaphor but it suggests a finish line. There isn't one. The retailers who are getting this right are treating change as a core competency instead of a project.
This connects directly to something we've written about: the difference between performance and transformation as goals. They require fundamentally different things from your organisation.
Governance is the bottleneck in most retail tech programmes. If every call needs a committee, you've already lost the pace battle. The organisations moving quickly have given their people genuine autonomy, and the accountability that comes with it.
We heard similar at last month's Retail Hive. The warning signs trace back to governance gaps and slow decisions, not technical failures.
Last year's sceptics are this year's advocates. The conversation has shifted from 'should we?' to 'how do we govern it safely?' It's good progress.
Use cases being actively deployed right now include price scraping, onboarding, data analysis and operational tooling.
The question being asked more carefully now is where the boundaries are when it comes to governance frameworks, audit trails and transparency to customers.
The opacity question is one worth taking seriously. Our piece on the seductive peril of the black box covers exactly why 'it works' isn't sufficient justification on its own.
Deprecation cycles are fast. Costs vary significantly by model. Dependency risk is real. The advice coming from operators who've already been stung: build flexibility in from the start.
The architecture decisions you make today will determine how much freedom you have in 18 months.
This came up repeatedly and deserves more airtime than it usually gets. Younger team members are genuinely worried about what AI means for their careers. The right framing is understood at the top but isn't filtering down consistently.
Leaders need to say it out loud, and keep saying it. Silence on this doesn't read as reassurance. It reads as confirmation of the fear.
There's a broader point here about the end of a certain model of staffing technology programmes. Our piece on the end of bodyshopping is relevant context.
And very few system integrators can actually do it well. The front end gets the attention; the back end is where programmes succeed or fail.
Know exactly what you're buying before you start. Ask direct questions about back-end capability and ask for references specifically for that work.
Covering the same dynamic, we've written about the problem with SIs marking their own homework.
Every app added post-launch compounds complexity and cost. Total cost of ownership on Shopify implementations is routinely underestimated because the scope keeps moving. Be disciplined about what's in at go-live. Everything else is a future sprint.
For a broader view of what's lurking beneath the surface of retail tech stacks, our retail technology iceberg piece maps the full landscape.
'Test the hell out of it.' Said on stage about Shopify, applies to any major re-platform. What we heard reinforced what we already know: 'tested' means very different things to different people. Integrations, third-party dependencies, data flows – these are the parts that break first in production and they're frequently assumed to work rather than verified.
Full end-to-end testing is about risk as well as confidence. Do you know where your risk actually sits?
Particularly in scenarios where you can't capture data from everyone in a booking party. The points mechanic is well understood but the harder question is around how to build genuine loyalty when you don't have a direct relationship with every person who benefits from it. Watch this space.
Be explicit about which one you're pursuing. They need different people, different processes, and different measures of success. Confusing the two, or trying to do both simultaneously with the same team, is one of the most common ways retail tech programmes go wrong.
Across two days, a few clear themes emerged.
The industry's relationship with AI has matured faster than most predicted. The governance conversation is now live and that's healthy.
On transformation, the gap between ambition and execution capability remains significant. Building the organisational muscle to change continuously is harder – and more important – than any individual programme.
On platforms like Shopify, the complexity hiding in the back end and the post-launch scope creep are predictable problems. They're also preventable ones, if you go in with clear eyes.
De-risking complex retail tech stacks
What retail tech leaders are really talking about: March's Retail Hive round tables
Marking their own homework – 30 years of SIs being paid twice
Do you know where your risk actually sits?
The Seductive Peril of the Black Box: Navigating Transparency in the Age of AI
The End of Bodyshopping: Farewell to the Great Desk-Warming Economy
How to move to continuous delivery in retail without increasing risk