Retail AI is usually introduced through the most visible outcome: a customer assistant, an in-store avatar, an automated forecast. Esselunga is starting somewhere much less photogenic — with the people who will have to use the technology every day.
In its 2025 consolidated financial statements, the Italian retail group said it had launched a pilot using AI tools, particularly ChatGPT Enterprise. A group of key users was asked to test the technology in real working environments, identify concrete use cases and establish an initial set of practices that could be replicated elsewhere in the organization.
In 2026, the pilot is being paired with a multi-stage training program: an initial assessment, AI-literacy webinars, “AI for Work” sessions tailored to different skill levels, and follow-up work to gather experience and feedback.
It does not sound revolutionary. That is precisely why it is interesting.
Enterprise AI programs often fail for reasons that have little to do with the model. The organization does not know which process should change, which data can safely be used, who remains accountable for the output or how the benefit should be measured. Buying licenses is much easier than changing a way of working.
Esselunga is treating AI adoption as an organizational capability rather than a feature launch.
That does not guarantee a financial return. The company still has to show which use cases generate measurable savings, speed or quality improvements. Many experiments may never get beyond the pilot stage.
But there is a useful lesson here for Italian retail and beyond: the most important part of an AI deployment may not be the model the company chooses.
It may be the ability to turn one successful test into a repeatable practice without losing control, quality or accountability.
Before retailers can hand processes to agents, they need to know which decisions are worth delegating in the first place. Esselunga is starting there.