When retail companies talk about AI, the default image is usually a shopping assistant recommending products to customers. Mercadona’s example is much less visible — and that is exactly what makes it interesting.
According to an account from the retailer’s technology team reported by Xataka, Mercadona decided to replace the external search engine it had used for years with an internally developed system. An initial prototype was built with help from Claude Code over a long weekend, and the project reached production roughly a month later.
Mercadona’s online store processes millions of searches each week, so this was not an internal demo. The new search system had to handle real traffic, understand the retailer’s catalog and work inside infrastructure already carrying a live business.
The eye-catching number is the development speed. But reducing the story to “AI wrote thousands of lines of code” misses the point.
The engineers still made the consequential decisions: architecture, validation, design and technical criteria. AI reduced mechanical work and made iteration faster. It did not remove the need for judgment.
That may be a more immediately useful form of generative AI for large retailers than another customer-facing chatbot.
Retail companies accumulate old systems, custom integrations and projects whose cost is often less about inventing an algorithm than about writing, testing and maintaining the software that connects everything together. If tools such as Claude Code let small teams rebuild strategically important components in weeks rather than months, the economics of build versus buy begin to change.
The risk changes too.
Generating software faster can also generate technical debt faster. An internally owned search engine still needs monitoring, relevance tuning, performance work, security and long-term maintenance. Building it quickly does not mean owning it will stay cheap.
That is why Mercadona’s case is useful. It does not show that AI can replace an engineering team.
It shows something more believable: AI can materially increase the amount of software a good engineering team can ship.
Retail’s AI shift may not begin with a shopper talking to a bot. It may begin when two or three engineers can build something that previously required half a team.