AI Rating in retail: the gate is workforce systems
Decent Readiness and Delivery, Risk collapsing for one reason: predictive modules inside workforce management software, never classified as Annex III systems.
In short. In retail the AI maturity profile carries an anomaly compared to other sectors: organisations score decently on Readiness and Delivery, and collapse on Risk for one reason alone, the systems managing staff. They arrived inside workforce management software, nobody classified them, and they fall under Annex III. It is the most underestimated gate we come across.
Chains arrive at the rating convinced the AI Act barely concerns them, because few commercial use cases land in the high-risk bracket and the sector narrative has pushed that point for months. The assessment moves attention somewhere they didn't expect, onto employees rather than customers. It changes the conversation.
The four dimensions in retail
Readiness measures preparation, data, skills, infrastructure. It is generally decent, because sales data exists, is structured and already in use. The weak point concerns undigitised store data and the split between physical and online channels, which in practice means the data on a single customer lives in at least three different systems, the till, e-commerce and the loyalty programme, without a single view a model could work on without manual reconciliation.
Delivery measures the ability to reach production and maintain it. In retail it is limited by the distance between head office, where systems are chosen, and the network, where they are used. A system released without support in store gets worked around within weeks, because staff find a shortcut that works better than the intended process and nobody at head office notices until the numbers stop adding up.
Risk measures governance, compliance and ethics. It is the dimension that sets the class, and it falls almost always on the same point.
Confidence measures management commitment and user trust. High at head office, low in store, with a gap wider in retail than elsewhere because the two populations rarely talk to each other.
The gate: workforce systems
The recurring constraint concerns shift planning, task allocation and performance evaluation systems. These have existed in retail for years, were bought as administrative modules, and in recent years acquired predictive features without the purchase being reassessed.
They fall under Annex III, because they concern the management of workers. Until they are classified as such, risk assessment, documentation and oversight are missing, and the Risk dimension stays below threshold whatever the quality of governance elsewhere.
There is an Italian aggravating factor. Labour law constraints and consultation with worker representatives sit on top of those systems, and a company that introduced them without that step has a problem documentation alone won't solve.
Video analytics, the point that surfaces late
Many chains run video surveillance installations that over time acquired analytical functions: footfall counting, heat maps, in some cases recognition.
The line between aggregate analysis and biometric processing is thinner than people assume, and crossing it brings heavy constraints. In assessments this surfaces late, because the installation is managed by physical security and never enters the IT perimeter.
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 RatingWhat a class means in practice
| Class | Typical situation in retail | What is reasonable to do |
|---|---|---|
| D | AI modules live inside software, no mapping, staff systems unclassified | Map the systems starting from those touching employees |
| C | Commercial inventory in place, HR systems unclassified | Classify workforce systems and open the internal conversation |
| B | Classification complete, oversight on Annex III systems | Extend documentation and review video analytics |
| A | Mature system verified across all dimensions | Reserved for verified assessments, not self-assessments |
Moving from C to B in retail is technical and relational at once, because it touches the organisation of work. It is also why it takes longer than management expects.
Why measure before the next software renewal
In retail the useful moment coincides with the renewal or replacement of workforce management or warehouse software.
That is when the supplier can be asked for documentation on predictive modules, a description of adjustable parameters and access to logs. After signature those requests become renegotiations, and the supplier often replies that it doesn't produce that information.
A chain renewing without having run the assessment carries an unclassified system for the whole term of the contract.
Where to start
The prerequisite is the AI system inventory, built starting from the systems touching employees rather than commercial ones.
The full model, with the four dimensions and the gate logic, is on the AI Rating page. The sector's regulatory picture is on the AI governance for retail page.
For an assessment of your chain's position you can book a session or start the self-assessment.