Step 01
Free diagnostic
We look at your real volumes per channel, the breakdown of contact reasons, the time to first response and what your customer databases actually contain in the light of Law 09-08.
Sector — Distribution
The customer writes. They won't wait
AI in Moroccan retail: multichannel customer service, product listings, forecasting, customer insight. Personal data, Law 09-08 and the CNDP.
In brief
In Moroccan retail, the customer messages on WhatsApp, in Darija, in the evening, and expects an immediate reply. Generative AI answers that pressure on customer service, product listings, forecasting and the analysis of customer feedback. The point to watch is personal data: loyalty programmes and marketing databases fall under Law 09-08 and the oversight of the CNDP, consent included.
Moroccan retail has changed pace without its tools keeping up. The customer compares online, writes on WhatsApp or on social channels, often in Darija, sometimes at ten at night, and considers a reply the next morning to be no reply at all. Retailers have responded by stacking up channels; they have not solved the underlying problem, which is one of volume and of language.
Generative AI addresses exactly that problem. It reads the message as written, retrieves the right information from the retailer's catalogue and procedures, proposes an answer, and lets the adviser decide on the cases that matter. In the background it also tackles what is expensive and invisible: the thousands of product listings to write, the customer feedback that goes unused, the forecasts made by guesswork three weeks before Ramadan.
Updated 14 July 2026
Sector stakes in Morocco
Retail: your customer data is your asset — and your exposure
A retailer accumulates personal data at a pace it rarely measures: loyalty card, purchase history, phone number, delivery address, exchanges across messaging channels. These processing activities fall under Law 09-08 and the oversight of the CNDP, with two points of friction that are constant in retail. The first is purpose: a database built for loyalty cannot be freely repurposed for training or marketing targeting without checking what the customer actually consented to. The second is prospecting: direct commercial solicitation requires consent, and plugging an AI tool into a contact list does not create that consent retroactively.
On top of that comes a linguistic reality most imported solutions ignore: your customers write in Darija, often in Latin characters, with abbreviations and a mix of French and Arabic. A model that does not understand this register will produce off-topic answers, and the adviser will redo everything — the gain vanishes. We therefore evaluate models systematically on your real, anonymised conversations before recommending an architecture. It is a test few vendors agree to sit, and it is a knockout criterion.
Purpose of loyalty data
A database built for loyalty cannot be freely used for other ends. Law 09-08 requires checking the declared purpose and the consent obtained before connecting any tool to it.
Prospecting and consent
Direct commercial solicitation requires prior consent. An AI assistant changes nothing here: it industrialises the message, it does not create the right to send it.
Darija and language mixing
Your customers write in Darija, in Latin characters, mixing French and Arabic. A model that cannot hold this register costs advisers time instead of giving it back to them.
Use cases
What AI concretely changes — distribution
Every use case links to the Hunter BI offer that delivers it. We claim no result figures until they are measured at your organisation.
Sovereignty
Sovereignty in retail: a question of the customer file
A retailer has no industrial secret in the sense a factory does, but it holds something just as coveted: a customer file. Hundreds of thousands of people, with their habits, their contact details and their purchase history. That file is an asset — it is also a responsibility under Law 09-08, and a target. Plugging that asset into an AI service without knowing where the data is processed, how long it stays there and who can access it means moving the risk out of your sight without having reduced it.
In retail, the architecture we recommend is almost always hybrid, and the dividing line is clear: processing that touches named-person data — segmentation, history analysis, customer conversations — happens within a controlled perimeter, on your premises or with a qualified host in Morocco; tasks with no personal data — product-listing generation, competitive monitoring, marketing copy — rely without difficulty on enterprise cloud platforms. This separation costs little to put in place at the outset, and it is almost impossible to retrofit after the fact.
Where to start
Three steps, in this order
In retail, proof comes fast: one channel, one team, two weeks of real conversations — and the debate about Darija is settled by the facts.
Step 02
Pilot on one channel
One channel, one team, one metric measured beforehand: time to first response, adviser hand-off rate, satisfaction. Models are tested on your conversations, not on a generic demo.
Step 03
Rollout to the stores
Once customer service is validated, the assistant extends to the network — aisle managers, franchisees, back office — as a per-seat subscription, with continuous supervision of answer quality.
Frequently asked
Does the AI really understand our customers' Darija?
Recent models do far better than they did two years ago, but quality varies sharply by model and by register — Darija in Latin characters, abbreviations, mixing with French. This is a question settled by testing, on your anonymised conversations, before any commitment. We refuse to recommend a model on this point without having measured it on your own data.
Can we use our loyalty database to personalise responses?
Within the limits of the declared purpose and the consent obtained. A database built for a loyalty programme cannot be freely repurposed for training or targeting. The question needs to be raised before connecting the tool, with your compliance officer — and it has answers, provided they are worked through at the right time.
Can an assistant answer on its own, with no adviser behind it?
On simple, documented questions — opening hours, availability, order tracking, return procedure — yes, with supervision. On disputes, refunds and anything that commits the retailer, no: the request is routed to an adviser with the history already summarised. The right setup is not the one that answers everything; it is the one that knows when to stop.
Do you need an e-commerce site to benefit from AI?
No. Purely physical retailers have two immediate opportunities: customer service on messaging apps — which already exists, even without e-commerce — and support for in-store teams. The catalogue and product listings become a priority as soon as an online channel opens, but they are not the mandatory starting point.
AI in your field: let's start with a diagnostic.
A free scoping session with a Hunter BI consultant: your data, your regulatory framework, the use cases worth launching first — and the ones better set aside.
- Réponse sous 24 h ouvrées, par un ingénieur
- Diagnostic gratuit, sans engagement
- Membre des réseaux partenaires OpenAI et Anthropic