AI Needs Somewhere to Go.

The real value of retail AI is not in the insight. It’s in what the retailer does next.

The retail industry is rapidly adopting AI, but adoption alone does not guarantee meaningful change. We have moved incredibly quickly from asking “What could AI do?” to embedding it into almost every part of the retail technology stack. Forecasting, customer segmentation, content generation, reporting, search, personalisation, pricing and analytics are all being reshaped by AI, and there is certainly no shortage of experimentation.

I’m not convinced retailers have ever really had an insight problem. Most retailers are already surrounded by a lot of information. They have sales reports, dashboards, customer data, promotional results, forecasts, campaign plans and mountains of operational reporting. They are not short of ideas either. The difficult part has always been what happens next: can the retailer actually act on what they know?

We often see retailers almost hamstrung by their own insights and dashboards. They know something needs to change. They can see that a category is underperforming, a campaign isn’t landing or one group of stores is behaving differently from another. What is much harder is understanding why, deciding what to do about it, and then what is even harder is turning that decision into action across a complex retail network.

That is where I think the next phase of retail AI gets much more interesting.

AI needs somewhere to go

An insight that ends in a dashboard may be useful, but ultimately nothing has changed. An insight that changes a range, removes unnecessary work, alters a planogram, improves a process, changes a store task or leads to a better commercial decision has created something tangible.

The opportunity now is to move beyond AI that tells us something towards AI that helps us do something.

Before intelligence comes foundations

There is a temptation to believe retailers need pristine data, perfect processes and a completely integrated technology stack before they can properly embrace AI. They don’t, and realistically, most retailers probably never will. Retail simply has too many moving parts. Teams are constantly sprinting the marathon of day-to-day operations, while products, prices, promotions, customers and priorities continue to change around them. Data will often be muddied, incomplete or still a work in progress.

But that shouldn’t become an excuse not to start. However, it does mean retailers need to understand the environment they are asking AI to operate within.

For physical retail in particular, that foundation includes some very practical information. What exists across the store network? What is the capacity of each location? What fixtures, digital touchpoints and merchandising spaces are available? What ranges can physically fit? What is each store already being asked to execute?

These may not sound like exciting AI questions, but without those foundations it becomes difficult to translate intelligence into something useful at store level.

Automating a broken process does not magically fix it either. In fact, one of the useful side effects of introducing automation is that it often exposes the shadow processes, spreadsheets, workarounds, inconsistent data and duplicated effort that have quietly become part of the way a business operates.

This can be very uncomfortable and embarrassing for most retailers to see, but it can also be incredibly valuable. You don’t need to wait for perfection. You need enough structure to begin, and a willingness to improve the foundations as you go. Just start!

Make AI useful before making it clever

This is also why some of the most valuable applications of AI in retail may not initially look particularly revolutionary.

AI and automation can remove the work that doesn’t require human judgement: tailoring instructions to individual stores, generating store-specific visual merchandising guides based on capacity or range, prioritising tasks according to commercial importance, or validating whether a campaign can realistically be executed before it reaches the frontline.

None of this comes with the theatre sometimes associated with AI. What it does is remove friction.

It reduces repetitive manual work and interpretation. It can give frontline teams clearer, more relevant instructions and give support-office teams greater confidence that what they are asking stores to do can actually be delivered. Most importantly, it frees people to spend more time on the things humans are considerably better at: judgement, creativity, problem-solving and serving customers.

Perhaps that is one of the simplest tests retailers should apply to AI. Has it made something easier? Has it removed unnecessary work or ambiguity? Has it helped somebody make a better decision or created more time to spend somewhere valuable?

If so, it is already doing something worthwhile.

Visibility is valuable. Action is better.

For years, retailers have invested heavily in getting better visibility of their businesses, and rightly so. Knowing what happened, where it happened and when it happened is incredibly useful. AI will make this even faster by surfacing patterns and anomalies that humans may struggle to find amongst huge volumes of data.

But visibility is not the destination.

The next step is using that information to decide what should change. Instead of knowing only whether a merchandising task was completed, imagine understanding the level of effort it took to execute across an entire network and whether the commercial result justified that effort. Instead of simply celebrating a lift in campaign sales, retailers should increasingly be able to understand whether the incremental profit justified the cost, labour and operational complexity required to achieve it.

The questions become more meaningful. Not just “What happened?”, but “Was it worth doing?” and, ultimately, “Should we do it again?”

Sometimes the most valuable recommendation AI could give a retailer might simply be: stop doing this, it isn’t worth it.

It isn’t particularly futuristic, but it is useful, and something retailers don’t often get enough time to reflect on as they are already knee deep in repeating the same campaign.

Closing the loop

This is where I believe the real opportunity for AI in retail sits: closing the gap between planning, execution and performance.

Many retail processes still operate in a reasonably linear fashion. We plan something, execute it, review the result and move on to the next thing. The future should be far more connected: plan, execute, measure, learn and use those learnings to improve the next decision.

If a store, a fixture, a category or a product underperforms, telling the retailer it underperformed is only the beginning. Why did it happen? Was the range wrong? Was stock unavailable? Was the execution poor? Was too much space allocated? Did the campaign arrive too late? Or was the store being asked to execute something that simply did not make commercial sense in that location? Or did the cost of that campaign outweigh the actual revenue?

This is where AI needs somewhere to go.

The intelligence should feed back into an action. Change the range. Change the space allocation. Adjust the planogram. Remove a task. Change the timing. Or give one store a different instruction from another because their customers, inventory and performance tell us that they should not be treated exactly the same.

That is the difference between AI sitting beside the retail operating model and becoming part of it.

Start with the problem, not the AI

So perhaps the question for retailers should not be “Where can we use AI?” Start instead with the problems that already exist in the business.

Where is friction slowing people down? Which decisions are consistently difficult to make? Where are teams repeatedly doing manual work? What information is already being collected but not acted on? What are you measuring without really learning from?

And, most importantly: what action do you want to become smarter?

Then work backwards into the technology.

Perhaps the next phase of retail AI isn’t about discovering more at all. Retailers have spent years building data lakes, dashboards, reporting suites and increasingly sophisticated analytics. There is already an extraordinary amount of information available, and AI will undoubtedly make those insights easier and faster to surface.

But the retailers that gain the greatest advantage may not be those with the most AI or the cleverest dashboard. They may simply be the ones that have built the ability to act on what it tells them.

Because the value of an insight isn’t the insight. It’s what changes next.