Back to blogStrategia

    AI adoption in retail: the use case that pays and the one that exposes

    Demand forecasting is measurable in four weeks and stays at minimal risk. Personalised pricing brings disclosure obligations and a reputational cost that rarely enters the business case.

    ZeroFive.AI May 22, 2026Updated on September 23, 2026 5 min

    In short. In retail the use case with the best ratio of value to friction is demand forecasting, because it produces measurable effects within weeks and stays at minimal risk. The case that attracts most attention is personalised pricing, which works technically and carries disclosure obligations towards the consumer plus a reputational risk that rarely enters the business case. Workforce systems, the highest-risk ones, get adopted without anyone treating them as such.

    Margin pressure in retail makes every percentage point interesting, and that pushes organisations to see AI as an optimisation lever rather than a system to govern. The consequence is that adoption comes first and classification later, or never.

    Where results show immediately

    Retail has an advantage other sectors lack: the cycles are short. A demand forecasting model is judged on four weeks of sales, not on three years of claims or two sporting seasons.

    That changes the logic of adoption. A chain can afford to get the first use case wrong, as long as it measured beforehand, because the cost of the error is recovered quickly. It is the opposite of what happens in manufacturing or insurance.

    The condition is that a baseline exists. Without the starting figures on stockouts, unsold inventory and margin, a four-week assessment produces an impression rather than a number.

    The map of candidate cases

    Use caseValueRegulatory frictionSuitable as a first
    Demand forecasting and replenishmentHigh and quick to measureLowYes
    Space and assortment optimisationMedium-highLowYes
    Customer service assistantMediumLow, Article 50Yes
    Online product recommendationMedium-highLow, GDPR profilingYes
    Personalised pricing per customerHighMedium, disclosure and reputationAssess carefully
    In-store video analytics with recognitionMediumHigh, biometric areaNo
    Shift planning and staff evaluationMediumHigh, Annex IIITreat as a project of its own

    The last row is the one adopted with the least caution in retail, because it arrives inside workforce management software and looks like an administrative feature.

    The personalised price case

    It deserves isolating because it generates confusion. A system adapting the price to an individual customer on behavioural data is not prohibited, and it carries a disclosure obligation towards the consumer: personalisation based on automated decisions has to be communicated.

    The delicate part isn't regulatory. In retail the perception of different treatment between customers produces reactions out of proportion to the size of the difference, and those reactions become a real commercial cost.

    The decision belongs in the open, weighing expected margin against exposure, rather than surfacing when somebody compares two screens.

    Which framework does your company actually need?

    AI Rating measures maturity across the four areas of the model and shows where to start, with priorities and estimated effort.

    Start your AI Rating

    How many customers the chain actually recognises

    Personalisation, in any form, only produces effects on the share of customers the chain recognises, and in physical retail that share is almost always a minority of receipts, because outside the loyalty programme the transaction stays anonymous and the model works on a sample nobody chose. A chain with thirty per cent of receipts identified is building predictions on a third of its customers.

    Before funding a personalisation use case it is worth measuring two numbers: the percentage of transactions linked to an identified customer, and the percentage of those customers reachable with valid consent. The second is almost always lower than the first.

    If both are low, the first budget pays off better spent on the reasons a customer should identify themselves at the till, because without that base any personalisation model works on a share too small to move the margin.

    When the pilot doesn't transfer to the network

    The right perimeter in retail is a group of stores or a product category, with the baseline measured on that same perimeter.

    One caution is specific here. Stores differ by catchment, floor space, assortment and customer mix, and a result obtained on ten selected stores does not transfer automatically onto five hundred. Sample selection counts as much as the model, and a sample drawn from the best-performing stores produces a result that won't repeat.

    The exit plan concerns integration with till systems and warehouse management software, which is the longest part and the one rarely included in the initial estimate.

    Who answers for the system

    In retail, ownership tends to land in whichever function bought the tool, often marketing or IT. For systems touching assortment the right point is the category manager, for those touching staff it is the HR director, and in that second case responsibility cannot be delegated to a supplier.

    One clarification belongs here. A system assigning shifts affects working conditions, and the person answering for it has to be able to change the parameters and stop the system, not merely receive its output.

    Where to start

    The sequence starts from the AI system inventory, which in retail has to include AI modules switched on inside software already in use, starting with workforce management.

    The general selection criteria are in AI adoption: how to choose the first use case, the cost model in the budget of an AI project.

    The sector's regulatory picture is on the AI governance for retail page.

    To assess which use cases make sense in your chain you can book a session.

    Want to discuss this for your company?

    30 minutes with us to figure out where to start, or an AI Rating to measure your starting point.

    #AI adoption#retail#distribution#pricing#Annex III
    Share

    Keep reading