The verdict in three sentences
A RAG assistant (retrieval-augmented generation) answers from your own documents, offers, procedures and ticket history rather than from the model's general knowledge: that is what makes it reliable for a telecom operator or ISP. For a support team receiving 4,000 requests a day, the project costs 6 to 18 million FCFA (about 9,150 to 27,450 EUR) to set up and 0.5 to 3 FCFA per message in API fees, for an autonomous resolution rate of 40 to 60%. Profitability depends less on the AI model than on the quality of the knowledge base and the escalation rule to a human agent.
What an ISP's customers in Dakar actually ask
Before pricing a chatbot, you need to classify requests. At a Senegalese ISP, a large share concerns information already available in internal documents or in the billing system.
| Contact reason (ISP in Dakar, 2026 estimate) | Share of requests | Possible autonomous RAG resolution |
|---|---|---|
| Outage or slow connection | 25 to 30% | 30 to 40% (guided diagnosis, ongoing incident) |
| Bill, balance, Wave or Orange Money payment | 20 to 25% | 60 to 75% (with customer account access) |
| Offers, prices, plan change | 15 to 20% | 70 to 85% |
| Installation, technician appointment | 10 to 15% | 40 to 50% (appointment booking) |
| Router setup, Wi-Fi, password | 10 to 12% | 65 to 80% |
| Complaint, cancellation, dispute | 5 to 8% | 0 to 10% (mandatory escalation) |
| Requests in Wolof or Wolof-French mix | 20 to 35% of total | 50 to 70% depending on model |
A weighted average of 40 to 60% resolution without an agent is therefore realistic, provided the assistant is connected to billing and network monitoring.
How much a RAG assistant costs in FCFA
The budget splits between preparing the knowledge base, channel integration (WhatsApp Business Platform, website, app) and connections to the information system.
| Item | Essential version | Full version | Comment |
|---|---|---|---|
| Request audit and knowledge base writing | 1,500,000 FCFA | 3,500,000 FCFA | 150 to 400 answer sheets |
| RAG engine (indexing, retrieval, guardrails) | 2,000,000 FCFA | 5,000,000 FCFA | cloud or local hosting |
| WhatsApp Business and website integration | 1,000,000 FCFA | 2,500,000 FCFA | via Meta-approved provider |
| Billing, CRM, network monitoring connection | 1,000,000 FCFA | 5,000,000 FCFA | balance lookup, ongoing incidents |
| Agent console and escalation | 500,000 FCFA | 2,000,000 FCFA | history passed to agent |
| Total setup | 6,000,000 FCFA | 18,000,000 FCFA | 8 to 16 weeks |
| Model API cost | 0.5 to 1 FCFA per message | 1 to 3 FCFA per message | depending on model and length |
| Hosting and maintenance | 250,000 FCFA a month | 800,000 FCFA a month | knowledge base updates included |
On WhatsApp, replies sent within the 24-hour window opened by the customer are not charged by Meta in 2026; only business-initiated template messages (reminders, notifications) are, depending on category. With 4,000 requests a day and about 6 messages per conversation, API costs stay between 360,000 and 2,200,000 FCFA a month.
Escalation: the rule that protects customer satisfaction
A chatbot that insists loses customers. The assistant must hand over to an agent as soon as one of these signals appears: complaint or cancellation, two answers rated unhelpful, retrieval confidence below threshold, priority business customer, or an explicit request for a human. The agent receives the conversation summary and account details, saving 1 to 2 minutes per transferred contact. This is set up at design stage and measured weekly through the real resolution rate (customer who does not come back within 48 hours).
Mini case study
Need a professional website?
Kolonell builds websites that attract clients, optimized for the Sénégalese market. Free quote in 2 minutes.
Moussa, head of customer relations at an internet service provider in Dakar, handles 4,000 requests a day with a team where each agent processes about 80 conversations a day at a loaded cost of 350,000 FCFA a month. With a RAG assistant at 12,000,000 FCFA and 45% autonomous resolution, 1,800 conversations a day are absorbed. Moussa avoids 8 hires planned for the seasonal peak (2,800,000 FCFA a month) and reassigns the remaining capacity to level-2 support. Recurring costs: about 1,500,000 FCFA a month (API, hosting, maintenance). Net gain: 1,300,000 FCFA a month, payback in 9 to 10 months, with first response time cut from 25 minutes to under 30 seconds.
FAQ
What is the difference between a classic chatbot and a RAG assistant?
A classic chatbot follows fixed decision trees and resolves 10 to 20% of requests. A RAG assistant searches your documents for each question and reaches 40 to 60% autonomous resolution in telecom support.
Does the assistant understand Wolof?
The main 2026 models understand Wolof written in Latin script and Wolof-French mixes reasonably well, with answer quality at 50 to 70% of that obtained in French. A set of 50 to 100 real conversation examples clearly improves results.
Does customer data leave Senegal?
With a cloud API, yes, which requires a declaration to the CDP under Law 2008-12 and transfer clauses. Hosting the retrieval engine locally with an API model adds 1 to 3 million FCFA to the budget.
How long does it take to launch the assistant?
Count 8 to 16 weeks, including 3 to 4 weeks for the knowledge base and a 2-week pilot on 10% of WhatsApp traffic.
How do you measure success?
Track weekly the real resolution rate with no repeat contact within 48 hours, the escalation rate and post-conversation satisfaction. A realistic 3-month target: 45% resolution and satisfaction above 4 out of 5.
Let's scope your project. Share a one-month export of requests and your current channels: we will estimate your autonomous resolution rate, a budget of 6 to 18 million FCFA and a launch date. Detailed quote within 48 h. WhatsApp +221 77 596 93 33.
Mohamed Bah
Fondateur, Kolonell
Passionate about digital and entrepreneurship in Africa, Mohamed has been helping Sénégalese businesses with their digital transformation since 2020. Founder of Kolonell, he believes every SME deserves a professional and accessible online présence.
