Marks and Spencer Hands Its Product Data to AI, Betting That the Feed Is Now the Storefront

The Unglamorous Asset That Now Decides Sales

Marks and Spencer has deployed Lily AI's Product Intelligence Platform to automate and improve the creation of structured product data at scale, and while the announcement sounds like back-office plumbing, it sits at the center of how modern retail actually competes. Product data, the attributes that describe colour, fit, material, occasion and style, is the least glamorous asset a retailer owns and increasingly the most consequential. It is what search engines index, what recommendation systems reason over, and now what AI shopping agents read when they decide which products to surface. M&S is treating this unglamorous asset as a strategic one, and it is right to.

The problem Lily AI solves is that rich, structured product data is expensive and tedious to produce by hand. Retailers with vast catalogues have historically relied on sparse manufacturer descriptions and manual tagging that could not keep pace with the range or the demands of modern discovery. Lily AI's platform uses artificial intelligence to generate attributes at scale, enriching product feeds with the descriptive depth that algorithms reward. For M&S, the deployment turns a manual bottleneck into an automated pipeline, and the payoff shows up not in the warehouse but in the search results where purchase decisions are increasingly made.

The Feed Becomes the Storefront

The strategic thesis behind the deal is captured in a single line from Lily AI co-founder and chief executive Purva Gupta, who argues that in a world where your product feed is your storefront, the quality of your product content is a competitive advantage. That reframing is worth sitting with. For most of retail history, the storefront was a physical place and then a website, a curated environment the retailer controlled. In the emerging model, a growing share of customers never see that environment at all. They encounter products through search results, shopping comparison surfaces and, increasingly, AI assistants that read the feed and present a shortlist. The feed is what the customer meets first.

If the feed is the storefront, then the quality of product data is no longer a hygiene factor but a driver of revenue. A product described sparsely will be poorly matched to queries, ranked lower and shown less often, regardless of how good the product itself is. A product described richly, with the attributes that algorithms and agents care about, will surface more often to more relevant shoppers. M&S's head of online experience, Stephen Orford, said that Lily AI has become a core part of how the retailer manages and scales its product content, language that signals the platform is embedded in operations rather than bolted on as an experiment.

Measurable Lift, Not Just Better Hygiene

The results M&S reports move the conversation from theory to outcome. The retailer describes stronger product visibility, higher click-through rates and meaningful revenue lift across both paid and organic channels after deploying the platform. Those are the metrics that matter, because product-data projects have historically struggled to justify themselves against flashier investments in storefront design or marketing. When better structured data demonstrably lifts click-through and revenue, the business case stops being about tidiness and starts being about growth, which is the only argument that reliably wins budget.

The dual improvement across paid and organic channels is particularly telling. Paid gains suggest the enriched data is improving the efficiency of shopping ads, where relevance drives both placement and cost. Organic gains suggest the same data is lifting unpaid search visibility, a compounding benefit that does not carry a per-click cost. Together they indicate that product-data quality is a leverage point that improves multiple channels at once rather than a narrow optimisation. For a retailer of M&S's scale, small percentage improvements in visibility and conversion across an entire catalogue translate into material revenue, which is precisely why the unglamorous work of data enrichment is drawing serious investment.

Optimising for Machines, Not Just Humans

The deeper shift the M&S deployment represents is a change in audience. For decades, product content was written for human shoppers, prose designed to persuade a person reading a page. Increasingly it must be written for machines, the search algorithms, recommendation engines and AI agents that mediate discovery before a human ever engages. Those systems do not respond to persuasive copy. They respond to structured, comprehensive, accurately tagged attributes they can parse and reason over. Optimising product data for machine consumption is a different discipline from writing marketing copy, and it is the discipline Lily AI's platform is built to industrialise.

This is where the story connects to the broader wave of agentic commerce. As AI assistants take on more of the work of shopping, scanning catalogues, comparing options and assembling recommendations on a customer's behalf, the retailers whose data those agents can best understand will win placement. A shopping agent cannot recommend what it cannot accurately interpret. M&S is, in effect, preparing its catalogue for a future in which the most important reader of its product descriptions is not a person but an algorithm acting for one. That preparation, done early and at scale, is a quiet form of competitive positioning that will be hard for slower rivals to reverse.

What Retailers Should Take From It

The practical lesson for other retailers is that product data deserves to be treated as strategic infrastructure rather than as a clerical afterthought. The catalogues that are richly and accurately structured will be the ones that surface in search, in shopping comparison, and in the recommendations of AI agents. The catalogues that are not will become progressively invisible, no matter how strong the underlying products, as discovery shifts to channels that reward machine-readable depth. M&S has recognised that the economics of visibility now run through data quality, and it has industrialised the production of that data rather than leaving it to manual effort that cannot scale.

The broader strategic point is about timing. The retailers investing in product intelligence now are doing so ahead of the fullest expression of agentic commerce, building the data foundations before the agents that depend on them become the dominant path to purchase. That head start compounds, because enriched catalogues improve visibility today and position the retailer for tomorrow simultaneously. For business leaders in retail and consumer goods, the M&S deployment is a prompt to audit their own product data with a hard question in mind: when an AI agent reads our catalogue on a customer's behalf, will it understand us well enough to recommend us, or will it quietly pass us by.