AI in fashion retail: A Computer Weekly Downtime Upload podcast

Fashion retailer Asos has been working with Microsoft to fine-tune large language models (LLMs) specifically for fashion. “The longer you engage, the more context the stylist understands,” says Asos chief technology officer, Przemek Czarnecki. Czarnecki believes artificial intelligence (AI) is a technology that will completely change the way the business operates and how it offers services and products to customers. Looking at the Asos AI Stylist as an example, Czarnecki says: “AI is something that transforms the customer experience. It allows you not only to provide customers with a search that they're used to but it also allows customers to state their intent and discuss in plain language what they are looking for.” This is how AI can be used at the front-end to boost the customer experience. On the back-end, he says: “We are also using AI to make our business more performant, more operationally efficient.” He says AI is being used both to improve how Asos presents products to its customers and to augment teams. While it is easy to get carried away with the benefits of deploying AI, Czarnecki notes that the starting point for an AI initiative is not actually AI at all, but data. He says: “If you can't get your data right and the quality of the data across the enterprise is not sufficient then you your AI will have the wrong foundation.” The retailer previously had a handful of data platforms which were being used by different teams. As Asos embarked on its AI initiative Czarnecki says the first step was to centralise all the data into an Azure data lake. “Many companies struggle with having multiple sources of the same information,” says Czarnecki. “Bringing data into one central repository creates massive simplification.” That centralisation enables teams across Asos to access data in a self-service way, helping remove friction and accelerate decision-making, he says. Reducing complexity and improving operational efficiency not only cut costs but also enables Asos grow faster, says Czarnecki. Asos has deployed Microsoft M365 Copilot acrossd the business. “We rolled out a wealth of training programmes everybody has the same starting point,” says Czarnecki. “They can develop simple agents and start getting proficient with prompting, which sometimes is a skill that some people have to learn.” Beyond learning how to compose prompts for an LLM, he says Asos is also providing training to help its staff start developing agents. He says: “People will start developing agents for themselves and over time, we can see how these agents become useful.” The company is also working with Microsoft on Dynamics 365 Finance & Operations and AI agents to simplify and automate core back-office processes – such as invoice and inventory management, purchasing and so on – across finance and supply chain. Going forward, Czarnecki says Asos’ goal is to start connecting these agents so that they can perform process automation end-to-end across the business. As well as user-generated agented, more complex agents will be developed by central teams.” Asos has also worked with Microsoft to identify 93 potential agentic use cases.