Retail consists of work that repeats itself and work that revolves around the moment itself. Counting stock, adjusting prices, placing orders, processing returns, making schedules: that is a large part of the office and back-office hours, and it is work with a fixed structure. Opposite that stands the work at the counter, in the store, on the phone: someone torn between two sizes, a complaint about a delivery, a question that does not fit the script. These two types of work together drive the outcome of every AI ambition in this sector. Where the work is structured and the data is accurate, capacity shifts relatively quickly. Where the work revolves around reading a situation and bearing responsibility for a decision, little or nothing shifts.
The same three-way split that runs through all sectors applies here as well. Part of the work can be taken over by AI: price comparison, stock forecasting based on sales history, compiling standard reports, the initial sorting of return requests. Another part goes partway, with a human approving or rejecting with reason: a system proposes an order, a buyer checks it against seasonal influences or a supplier agreement the system does not know about. And a third part remains human work: the conversation with a customer who needs something specific, the judgment call in a warranty dispute, leading a team on the floor during a busy period. Which part of the work falls into which category differs per company. That depends on how standardized the assortment is, how clean the stock and sales data are, and how much of the customer contact already runs through fixed channels.
A retail chain with a limited assortment, central purchasing, and digital registers already has the foundation for a large part of the back-office tasks in place: the data is there, the process is repeatable, and a system can make a proposal that a human checks. A chain with a broad and changing assortment, decentralized purchasing per location, and many manual exceptions does not have that foundation. That difference is not in the ambition — both management teams can say the same sentence, "AI should take over our inventory planning" — but in what the organization can already do before that work can be shifted. Are the roles that now order manually also the roles that will assess a system proposal later? Has a decision-making authority been assigned to whoever handles the exception? Those questions lie beneath the ambition, and they become visible as soon as you test rather than assume.
If capacity is freed up because a task moves to a system with oversight, a question arises about what happens to those freed-up hours: other tasks, less hiring of temporary staff, a different staffing structure. That question is for the employer, and decisions affecting the workforce are subject to their own legal requirements. What is at issue here is the step before that: which work is, in terms of task structure and data quality, already at a point where a system can take it over or support it, and which work is not. That is a factual question about work, not advice about people.
Retail is not the only sector where this question plays out, and the answers differ by industry for comparable reasons. In hospitality the question runs differently because of direct service and variable staffing levels, as described in the AI ambitions at play in hospitality; in the agricultural sector, seasonal influences and physical work play a larger role, as described in how AI ambitions in the agricultural sector relate to seasonal work; and in financial services the emphasis lies with regulation and oversight, as elaborated in the AI ambitions within financial services and the role of compliance. The comparison shows that the ambition itself is less distinctive than the readiness behind it.
An ambition stating that AI takes over work presupposes capabilities that may not yet exist now: clean data, clear decision-making authority for exceptions, a team that can assess a system proposal rather than only execute it. Without that testing, the ambition remains a sentence on paper, a risk that plays out more broadly and is addressed in why a strategy sometimes remains a document instead of a change. Conversely, it is also possible for a change to already be taking hold before anyone gives it a name; which signals point to that is described in the early signals that a change is taking hold. The underlying question — which work in this specific company can genuinely be taken over by AI — is answered per task with the work scan from FTE TO AI.
The first step is not to reformulate the ambition, but to look at where the organization already stands. A free readiness check of eight short questions, one per dimension, gives a picture of where you are furthest along and where you are least far along. The full ambition test, with the four layers and the five confidence gates, is under construction.
Vertel wat u wilt bereiken, dan kijken we samen wat daarvoor moet staan.
Answers come from this site’s knowledge base. Not tailored advice, and not a scan of your company.