FAQ — AI agency & automation in Morocco

A production-grade AI project in Morocco runs 35,000–250,000 MAD depending on scope. A standalone AI agent (WhatsApp chatbot + knowledge base) starts at 35–70k MAD. A full marketing automation build (CRM + lead scoring + workflows) sits at 80–150k MAD. A custom SaaS replacement (internal, multi-user) exceeds 150k MAD. Atlaq never charges for discovery workshops or slides — only for production deliverables.

A WhatsApp agent connected to your catalog, availability and CRM costs 35–60k MAD to build, plus 800–2,500 MAD/month for hosting + LLM tokens (Meta Cloud API + OpenAI/Anthropic/Gemini depending on volume). For a restaurant, it typically pays back in 3 months from booking-management time savings alone.

A simple workflow (lead → CRM → email) costs 8–15k MAD to build. A business-grade orchestration (10–30 nodes, bank integrations, conditional logic, retries, alerting) costs 30–80k MAD. Self-hosted n8n runs ~150 MAD/month in infra; Make starts at $9/month; Zapier gets expensive past 2,000 tasks/month.

Yes — after a first production deliverable. Typical retainer: 25–60k MAD/month for continuous iteration, maintenance, AI agent monitoring, and new automations. We never sign a retainer without first shipping something measurable.

Simple AI agent: 2–3 weeks from signature to go-live. Full marketing automation: 4–6 weeks. Custom SaaS replacement: 8–14 weeks. We deliver in visible weekly sprints (live demo every Friday) — you never wait 3 months without seeing something work.

Yes, on a scoped perimeter: one AI agent + one integration + one channel (WhatsApp, web, or email). Beyond that, 2 weeks become theater. We'd rather ship a genuinely usable MVP than fake-ship a rushed full product.

Yes. Any project handling personal data goes through: prior CNDP declaration, hosting in an adequate jurisdiction (Morocco, EU, or adequacy-listed country), DPA signed with sub-processors (OpenAI, Anthropic, Google), explicit retention periods, and implemented access/erasure rights. We hand over the full compliance dossier at delivery.

Your choice: Morocco (N+1 datacenter in Casablanca via local partners), EU (Frankfurt/Paris on AWS/Hetzner), or hybrid. LLMs run in the EU (Azure OpenAI Sweden, Anthropic via AWS Frankfurt). For regulated sectors (banking, health, gov), we offer self-hosted open-source LLMs (Mistral, Llama) on dedicated GPUs.

No. We use enterprise API endpoints only (OpenAI API, Anthropic API, Azure OpenAI, Google Vertex), which do not train on your data by default. For sensitive workloads we use privately deployed models. None of our clients has ever had data leak into a public model.

Frontend: React/Next.js, Vite, Tailwind. Backend: Node.js, Python (FastAPI), Supabase, PostgreSQL. AI: OpenAI, Anthropic, Google Gemini, Mistral, self-hosted open-source (Llama, Qwen). Orchestration: n8n (self-hosted preferred), Make for simple cases. Vector DBs: pgvector, Pinecone, Qdrant. Hosting: AWS, Hetzner, Vercel, Cloudflare.

Yes — common integrations for our industrial and retail clients in Morocco. We use native APIs where they exist (Sage X3 Web Services, Odoo XML-RPC/JSON-RPC) or custom middleware for older versions. Typical timeline: 2–4 weeks for a clean bidirectional integration.

Yes. We handle automated bank reconciliation (CFONB/MT940 statements), Moroccan payment rails (CMI, PayZone, Naps, YouCan Pay for e-commerce), and open-banking where exposed. For enterprise flows we use secure SFTP + custom parsers when no public API exists — the most common case in Morocco.

No. We require a written technical brief (template provided) before any call. This eliminates 90% of wasted time on both sides and ensures the first real meeting is a paid scoping workshop (15k MAD, credited if we sign). Our job is to ship, not to pre-sell.

One of the two co-founders (Hamza Yajid or Mohamed Birouch) is technical lead on every project. No junior hidden behind an account manager. You get a direct Slack/WhatsApp channel with the technical team from day one to launch.

No. We ship clickable prototypes, video demos and technical docs. Zero PowerPoint. If you want a partner who spends 30% of their time in slideware, we're not the right agency.

We guarantee technical delivery (functional, performant, secure) and scope. We do not guarantee that AI will triple your sales — no honest operator can. But we systematically measure impact (conversion rate, time saved, cost/lead) and iterate on the numbers.

Across 12 tracked clients in 2025: average payback in 4.2 months for customer-service agents (WhatsApp + knowledge base), 6.8 months for full marketing automations. Outlier: one hospitality client paid back in 6 weeks from no-show reduction alone.

It works where data already exists (CRM, ticketing, ERP, catalog) and the process is repetitive. It doesn't magically work on dirty data, fuzzy processes, or teams that won't use the tool. Our first deliverable is always an honest audit: we tell you if your case is ready or if you need to clean data first.