Morocco's AI market filled up fast. Many players present an AI offering without ever having put a system into production. The difference does not show on a brochure — it shows when you ask for a URL.
4YA is an engineering firm founded in 2020, based in Marrakech, with operations in Casablanca. We design, build and operate our own products — they are live, you can open them right now. The same team handles client projects.
Our products, in production, publicly accessible
These are not demos. They run, they have users, and they were built with the same building blocks we deploy for clients: autonomous agents, computer vision, private LLMs, process automation.
How to spot a real AI engineering team
If you are shortlisting providers in Morocco, four questions are enough to sort them. None requires technical skill.
- "Show me a production system you operate yourselves." Not a mockup, not an NDA-covered client project: a URL. Many cannot answer.
- "Who writes the code, and is it subcontracted?" A subcontracting chain adds delay, dilutes accountability and complicates handover.
- "Where is my data processed?" If the answer is a US API, your data leaves the Kingdom — and law 09-08 / CNDP compliance becomes your problem, not theirs.
- "If I leave, what do I keep?" The code, models and data must be yours. Otherwise you are not renting a service, you are renting a dependency.
What we deploy for Moroccan companies
Autonomous AI agents and automation. Agents that execute complete business processes — follow-ups, qualification, document handling, internal support — built with LangGraph and n8n, deployed on your infrastructure. See the LangGraph guide for CIOs.
Computer vision and security. Real-time analysis of existing camera feeds for industry, logistics, retail and hospitality. See the remote monitoring guide.
Private LLMs and sovereign infrastructure. Open-weights models deployed on your infrastructure or locally, with no dependency on a foreign API. See the sovereign AI guide.
Bespoke engineering. Hardware and IoT, SaaS architecture, cloud infrastructure. See engineering mandates.
Sovereignty: a matter of architecture, not contract
Law 09-08 governs personal data processing in Morocco, under CNDP oversight. A contractual clause cannot guarantee compliance if processing happens abroad. We therefore favour local processing — on site or on Moroccan infrastructure — so the question stops arising: the data does not leave.
Where we operate
Head office in Marrakech, operations in Casablanca, and interventions across the Kingdom — Rabat, Tangier, Agadir, Fès and industrial zones. Software deployment is remote; hardware installation is on site.
Frequently asked questions
Which AI projects are actually mature in Morocco today?
Internal process automation with autonomous agents, real-time video analysis, till fraud detection, WhatsApp assistants in French and Darija, and automated document processing. These deliver measurable returns, not showcases.
Do you need to be a large company?
No. The most profitable deployments start small: one repetitive process, one site to watch, one WhatsApp channel to automate. Scale comes later.
Where to start an enterprise AI project
The best first project is not the most visible, it is the most measurable. Look for a repetitive, frequent process whose current cost is identifiable: handling inbound requests, document checks, follow-ups, manual re-keying between systems.
This criterion immediately rules out showcase projects. A conversational assistant on the public site is spectacular but hard to evaluate; automating an internal process is measured in hours saved from the first month.
What determines the budget
Three factors matter more than the rest: volume processed, error tolerance, and integration depth with existing systems. An agent reading documents and proposing an answer for validation costs far less than one allowed to act alone — because the latter requires guardrails, traceability and supervision.
Infrastructure also weighs in: hosting a model locally costs more upfront than an API call, but becomes cheaper at high volume — and settles the data residency question.
The most frequent mistakes
Starting from the tool. Picking a platform before scoping the process leads to bending the business to the tool. The reverse order yields better results and costs less.
Underestimating data. Most projects fail not on the model but on the quality and accessibility of internal data. A two-day data audit often prevents two months of drift.
No internal owner. An automated system needs a client-side referent able to validate edge cases and flag drift. Without that relay, even a technically successful deployment erodes.
4YA SARL AU — engineering firm founded in 2020 · Marrakech (HQ) & Casablanca · Operating across Morocco
Keywords: AI agency Morocco · artificial intelligence agency Morocco · custom AI agents · automation Morocco · WhatsApp chatbot Morocco · sovereign LLM