Artificial intelligence is on the agenda of every business conference in Casablanca and Rabat, but most Moroccan SMEs have a simpler question: what can it actually do for my business in the next few months? The answer is usually less spectacular than a chatbot demo, and far more useful.
Start with the process, not the AI
The biggest gains rarely come from AI alone. They come from removing manual steps between tools you already have: the email that becomes a spreadsheet row, the spreadsheet that becomes an invoice, the invoice that becomes an accounting entry. Automating those steps often needs no AI at all, just reliable connections between your tools.
AI is useful where the information is unstructured: scanned documents, free-text emails, voice notes, photos. That is where fixed rules alone cannot cope.
Five good first candidates
- Supplier invoices and receipts. Read the amounts, VAT, dates and supplier details, check them, and prepare the accounting entry.
- Customer requests arriving by WhatsApp and email. Sort them, pick out the order or complaint details, and send them to the right person with the context they need.
- Weekly and monthly reports. Gather figures from your sales, stock and finance tools into one report, automatically, every time.
- Matching orders, deliveries and payments. Compare them across platforms and flag only the exceptions.
- Internal knowledge. An assistant that answers staff questions from your own procedures, price lists and contracts, with a link to the source.
What is specific to Morocco
Languages
Moroccan businesses work in French, Modern Standard Arabic, Darija and often English, sometimes in the same document. Current AI models handle French and Arabic well. Darija, especially when written in Latin letters, is harder. Test on your real documents and messages in every language you receive, and keep a human review step for cases where the AI is unsure.
WhatsApp is a business channel
For many Moroccan companies, WhatsApp is where orders, questions and confirmations actually happen. An automation project that ignores it misses the main source of work. The WhatsApp Business Platform allows compliant, structured connections with your systems.
Data protection under Law 09-08
Personal data in Morocco is protected by Law 09-08 and overseen by the CNDP, which also regulates transfers of personal data outside the country. Before sending documents to any AI service:
- send only the data the task really needs;
- choose providers and settings that do not use your data for training;
- document the processing, and declare it where required;
- consider keeping the most sensitive processing on servers you control.
This is not legal advice, so involve your legal adviser. But designing for privacy from the start is far cheaper than adding it later.
Measure before you trust
An AI step should be treated like any other part of a system that can fail:
- Build a sample of real documents or messages with the correct answers. A few hundred is often enough to start.
- Measure accuracy for each field and each language before going live.
- Set a confidence threshold. Below it, a person reviews the result.
- Log every automated decision so errors can be traced and corrected.
- Track the cost per document, so the automation stays profitable as volumes grow.
A realistic roadmap
- Weeks 1 and 2: map how the process really works, choose one frequent task and collect a sample.
- Weeks 3 to 6: build the automation and the review screen, and measure accuracy on the sample.
- Weeks 7 to 10: run a pilot with one team, alongside the current process.
- Afterwards: reduce human review where accuracy holds, then move on to the next process.
The takeaway
AI automation in Morocco works best when it is unexciting: one process at a time, measured, reversible and designed for the languages and channels your business actually uses. Start small, prove the value on your own data, then expand.
Want to find out which of your processes is the best place to start? Talk to our automation team.