Chatbots & live chat
Some links on this page are affiliate links: we may earn a commission at no extra cost to you. Our approach
A chatbot is not there to replace salespeople or customer service. It is there to answer the questions that already arrive every day, at any hour, and to make sure no contact is lost outside office hours or during busy periods. ## Start from real questions The best chatbots are built from actual conversations: WhatsApp messages, emails, phone calls. If you know the ten questions your customers ask most often, you already know where the chatbot creates value. ## WhatsApp first in Morocco Many customers write, call or message on WhatsApp rather than filling in a form. A well-designed WhatsApp assistant captures these conversations and passes them on with the context needed. ## AI or rules Rule-based scenarios are predictable and cheap. An AI assistant answers open-ended questions but needs a validated knowledge base, supervision and fallback rules. ## Measure the real benefit A chatbot is worth it if it captures contacts, frees up time and escalates to a human at the right moment. Without regular review, it quickly becomes a source of frustration.When this guide is for you
- You receive repetitive questions (prices, opening hours, delivery) that take the team's time
- You sell B2B and need to qualify visitors before a salesperson travels to meet them
- You receive WhatsApp messages outside opening hours and lose them
- You want to capture contacts on the site without an overly heavy form
- Your catalogue or services need help with navigation
Step by step
- List the real questions — Analyse WhatsApp, email and call conversations to identify the 10 to 20 most frequent questions. The chatbot starts from these real questions, not from a theoretical list.
- Design the scenarios — Each scenario offers short answers, clear choices and a way out to a human. Also define what the chatbot must collect (name, phone, need) before handing over.
- Choose the platform and the AI — A rules-based chatbot is enough for simple answers. An AI assistant connected to your content handles more open-ended questions, but its cost and supervision are higher. The WhatsApp channel requires the Business API.
- Test and monitor — Test the scenarios with real questions, then monitor the conversations: resolution rate, enquiries handed to a human and contacts captured. Add answers to unanticipated questions.
What it costs (in MAD)
| Service | Indicative price | Good to know |
|---|---|---|
| Simple scenario-based chatbot (site) | 8,000 – 25,000 MAD (≈ 730 – 2,300 €) | 10 to 20 scenarios, capture form and integration |
| WhatsApp Business chatbot | 15,000 – 40,000 MAD (≈ 1,400 – 3,700 €) | WhatsApp API, accounts, scenarios and human handover |
| AI assistant based on your content | 25,000 – 80,000 MAD (≈ 2,300 – 7,300 €) | Knowledge base, models, supervision |
| Maintenance and improvement | 1,500 – 6,000 MAD / month (≈ 140 – 550 €) | Adding answers, analysing conversations |
| AI usage fees | 200 – 3,000 MAD / month (≈ 18 – 270 €) | Depends on conversation volume and the model |
Indicative ranges for the Moroccan market in 2026, excluding tax. Prices vary with scope and provider.
Mistakes to avoid
- Launching a chatbot that replies "I didn't understand" to half the questions, which annoys more than it helps
- Not planning a handover to a human for sensitive or complex cases
- Capturing personal data without explaining how it is used
- Ignoring conversations: without regular review, errors repeat and trust drops
Frequently asked questions
Does a chatbot replace customer service?
No. It absorbs simple questions and prepares the ground for humans. Complaints, negotiations and specific cases are still handled by people.
Should I start on the site or on WhatsApp?
WhatsApp is often more natural for Moroccan customers and captures more conversations. Start with the channel where you already receive messages, then extend to the site.
What is the risk of an AI chatbot?
It can invent information or promise an inaccurate price or deadline. Limit it to a validated knowledge base, with fallback answers and a human handover whenever it is unsure.
How many conversations can a chatbot handle?
Technical capacity is not the limit: what matters is the quality of the answers. A bot that handles fifteen frequent cases well is better than a bot that tries to handle everything.
How do I know whether it is useful?
Track the resolution rate without a human, the number of contacts captured and satisfaction after the conversation. A bot that hands over 90 % of conversations early is often a form in disguise.
Updated on
Related guides
Version française: /services/generation-de-leads-crm/chatbots/ · AR · NL