Separating the real from the hype
AI is everywhere in retail conversation, and most of it is noise. The question is not whether AI matters – it does – but where it is genuinely earning its keep today, and where it is still a solution looking for a problem.
This guide is deliberately practical. It focuses on the applications that are working now, the ones worth piloting, and the ones to avoid until they mature.
Where AI is earning its keep today
Demand forecasting – the most mature retail AI application. Modern forecasting handles more variables, more SKUs and more frequent updates than traditional models, and it is already improving stock accuracy and reducing waste.
Pricing and promotions – dynamic pricing, markdown optimisation and promotion targeting are delivering measurable margin improvements, particularly in categories with short shelf lives or volatile demand.
Personalisation at scale – recommendation engines and personalised marketing are now accessible to mid-sized retailers, not just the giants. The lift in repeat purchase and basket size is real.
Inventory and supply chain – AI-assisted replenishment and allocation are reducing out-of-stocks and improving stock turn, especially across multi-site estates.
Where AI is worth piloting
Visual merchandising support – tools that suggest planograms or flag display opportunities are improving fast and worth a controlled trial.
Customer service automation – well-designed chatbots and assistants are handling a growing share of routine queries, freeing people for the complex ones. The key word is well-designed.
Loss prevention – computer vision applied to CCTV is starting to deliver genuine shrink reduction, though privacy and trust must be handled carefully.
Where AI is not ready yet
Fully autonomous stores – the technology works, but the unit economics do not, outside a narrow set of formats and locations.
AI-generated product design – promising for inspiration and iteration, but not yet a replacement for human judgement in final design decisions.
Strategic decision-making – AI can inform strategy, but the idea that it will set your strategy is a distraction. Strategy is still a human discipline.
How to start
Start with a problem, not a technology. The retailers who benefit from AI are the ones who identify a specific, measurable challenge – forecasting accuracy, markdown depth, repeat purchase rate – and ask whether AI can do it better than what they have today. If the answer is yes, pilot it, measure it and scale what works.
The principle
AI in retail is a tool, not a strategy. Use it where it earns its keep, pilot it where the signal is strong, and ignore the rest until it is ready.