AI & the future of retail

AI demand forecasting combines sales per store and per day with the weather, bank holidays, promotions, local events and yesterday's sales. A supermarket learns that a warm Saturday calls for more watermelon and less soup; a fashion chain learns how many sizes of a coat each store needs. The system drafts the order, and the employee adjusts for what they know and the system does not. For fresh produce, forecasting matters even more. According to the UN Environment Programme, more than a billion tonnes of food were wasted worldwide in 2022, about 12 per cent of it in retail. The next step is a supply chain that steers itself: the system orders, allocates and marks down, and the employee steps in only when something is out of the ordinary.

With dynamic pricing, a retailer adjusts prices to demand, stock, shelf life or a competitor's price. Online shops have done this for years. Digital shelf labels bring it into the store: a central system pushes a new price to thousands of shelves at once. AI estimates which markdown makes it more likely that a product sells in time, such as meat that reaches its use-by date tomorrow or summer clothing at the end of the season. This is not personal pricing: here, everyone in the same shop at the same moment pays the same price. Markdowns become a calculation per product and per hour, instead of a fixed round at closing time.

A digital sales assistant is a language model in the online shop or app that advises customers the way a good sales assistant does in store. The customer types or says what they are after: a rain jacket for an autumn bike ride, a present for a nine-year-old, a meat-free dinner for four. The assistant asks a follow-up question, compares products, explains the differences and puts a suggestion in the basket. With access to the range and live stock, it ties every recommendation to what can actually be delivered. Before long it will not wait to be asked: using data the customer chooses to share, it notices that the tyres are due for replacement and pairs the offer with a garage appointment.

Personalisation at scale links customer behaviour to range, offers and content. For each visitor, AI decides which products appear first and which offer they receive, based on past purchases, products viewed and the time of day. Someone who has just bought a tent sees sleeping bags, not a second tent. In a supermarket app, the usual shopping list is ready and waiting, with one new product that fits. Done well, it feels like a shopkeeper who knows you. The next step is an online shop rebuilt for every visitor, with images, copy and an offer that exist for that one person alone.

In large online shops, product descriptions and product photos made by AI are now routine. Retailers are now going further, with entire campaigns. Generative AI creates images, videos and copy for each country, season and customer group, in the brand's own tone, and places a product in a fitting setting: a coat on a windswept quay, a sofa in a small city flat. The system then tests which version works and moves budget to whatever performs. A large fashion platform reports that most of the marketing content on its own site is already made with AI. The creative team sets the direction, guards the brand and decides what must never appear.

Smart in-store cameras turn images of shelves and queues into concrete tasks for the store team: an empty shelf, a product in the wrong place, a shelf price that does not match the till, a queue that is getting too long. Alerts arrive in an app on staff devices. The cameras look at shelves and queues, not at who the customer is. The next step is a camera that predicts rather than reports: it sees that the bread shelf will be empty within the hour, or that a rush at the tills is ten minutes away, and hands the task straight to a member of staff or a robot. The store team then works an hour ahead of the customer.

An agent is AI that independently takes the steps needed to finish a task. In a retail business, an agent can assess a return request, create the returns label and issue the refund. Another agent spots that a bestseller is running low, prepares an order with the supplier and adjusts the warehouse rota to match the delivery. Retailers are testing such agents now, usually starting with work that is frequent and low-risk. The next step is agents that coordinate with each other: the returns agent notices that a pair of trousers keeps coming back because it runs small, and the buying agent adjusts the next order.

A retail language model is a language model that works with one retailer's own knowledge: the range, the manuals, years of customer questions, the returns policy and the work instructions. Staff ask it questions through a handheld scanner or a headset. Will this floor paint work on concrete? Can a customer exchange an item without a receipt? How is a damaged delivery handled? The answer comes back in plain language, with the source attached. For a new part-timer, it is a colleague who is always available. Large retailers already use such assistants in their stores. The next step is a language model that sees through smart glasses and offers help unprompted, for example when a customer is looking for a product that is in stock at another branch.

In an AI-first retail organisation, AI and new technology take over routine work and forecasting: the work people do not need to do. Staff concentrate on what calls for human judgement, taste and contact with the customer. AI forecasts demand per store, reorders stock and proposes prices; an agent answers customer questions about delivery and returns. Staff assess the exceptions and talk through the options with a customer who is hesitating. The scenario for 2040: an organisation serves more customers with fewer people, because agents do most of the routine work.

When customers scan their own shopping, order online and let an agent pay, the work in retail changes. Till work shrinks, and roles appear that barely exist today. The supervisor of a team of agents decides what the buying, returns and customer service agents may do on their own, and follows their decisions. A personal shopper for regular customers uses the body model and preferences a customer shares with the shop to pick out what will suit them. And in a shop that restocks itself overnight, the daytime employee becomes a host: welcoming, advising and making a visit into something an online shop cannot offer. The knowledge of experienced sales staff finds a new place, in shops where AI handles more and more of the transactions.

Digital colleagues are AI systems and agents that take over a store team's recurring work. The day then starts with a list that is already waiting. Overnight, AI gathers the open tasks: which shelves are running empty, which products need marking down, which online orders must be ready by ten. It builds the rotas and, when someone calls in sick, proposes a new one. Agents place orders with suppliers, process payments and prepare campaigns for the shop, physical and online. The store manager approves or adjusts, because they know there is a street fair on Saturday. That frees up time for customers, offline and online: giving advice, sorting out a complaint, helping a child who has lost their parents.

Leadership in retail means looking ahead while keeping the shops and the online business running today. The change takes shape with the people who run the chain, the format or the branch today. It starts with initiative: making a start with AI and new technology in a few stores or in one process, and showing what it delivers. And with a picture of tomorrow's retailer that a store manager can retell to their team. That is how leaders get staff on the shop floor and at head office to help build it.

Every organisation has people who have worked there for years and people who have just started. Both groups need to learn to work with AI and new technology. In a future academy, the organisation arranges this itself: with lessons at the moment a question comes up, and with programmes for leaders. It does not wait for schools and colleges to catch up. Everyone in the organisation learns tomorrow's work where the work actually happens.

A business increasingly delivers its service together with partners. Start-ups come up with new concepts, scientists test the results, sometimes from another country, and governments help agree shared rules on data, such as digital product passports. With the AI agent, a new player arrives: the agent that searches, compares and buys for the customer. Shared standards make services easier for those agents to find. Together, organisations also take on problems such as waste. Today's competitor is sometimes tomorrow's partner in a shared service.