Journal / Kirill Naumov
M&S Gives Store Managers AI. The Real Test Is the Morning Huddle
M&S announced 11,000 Copilot licences and AI training. A practical look at how store managers can turn summaries into better decisions on the shop floor.
By Kirill Naumov ·
A store manager does not need another beautiful summary at 8:45 a.m. She needs to know why the best-selling sandwich is missing, who can fix it before lunch, and what the team should tell customers meanwhile. That is the difference between giving a manager AI and making the store easier to run.
In March 2026, Marks & Spencer said it had bought 11,000 Microsoft 365 Copilot licences for every store manager and its support-centre colleagues. It also announced a training and development programme. M&S named meeting notes, sales insights, rotas and shift handovers as possible uses. These are the company’s plans and examples, not proof of time saved or improved store performance.
Make the handover answerable
A shift handover is useful when the next manager can act on it. A good AI-assisted version separates confirmed facts, unresolved questions and named actions. “Availability was poor yesterday” is a comment. “Five key lines were unavailable at noon; two have replenishment due at 10:00; check the other three with the stock team” is a handover.
The source matters. Before a summary goes into a morning briefing, a manager should be able to inspect the underlying stock or sales record and correct anything the tool has misunderstood. If the data is late, the summary should say so. A confident sentence built on yesterday’s inventory is worse than a short, honest warning.
Train for judgement, not prompt tricks
The training question is not whether managers can produce a polished paragraph. Give them a messy scenario: a promotion created a sales spike, a delivery is late, and two colleagues called in sick. Ask what the AI summary establishes, what it only suggests, what must be checked and what decision belongs to a human. Practise the conversation with the team, including how to explain a change of plan without passing off a guess as fact.
A second exercise should test refusal. What customer or colleague information should not go into an open prompt? Who is allowed to review a rota draft? When is a recommendation unfair or simply impractical? A manager who spots a wrong assumption is using the tool well, even if the final answer is “ignore this suggestion”.
Measure a better shift
M&S says the goal is to free managers for customers and colleagues. To test that promise, compare similar stores or shifts before and after adoption. Track time spent preparing handovers, time on the floor, recurring stock issues, corrections to AI summaries and whether the next shift actually completed the named actions. Watch for a hidden cost: if support-centre staff spend hours cleaning input data or correcting outputs, the apparent saving in the store is incomplete.
The lesson for any retailer is not to buy 11,000 licences. Start with one decision a manager makes every day, give it reliable inputs and a clear owner, then train people to challenge the output. The best AI morning huddle is the one after which the team knows what to do.
Explore the English AI for Retail workshop → · Discuss AI training for store managers
Primary source: M&S’s March 2026 Copilot announcement. The rollout and intended uses are company-reported. The training and measurement suggestions above are independent proposals, not a client project or measured results.