WhatsApp chatbot for SMEs in Benin: what it actually does (and what it costs)
A customer messages on WhatsApp at 10pm. They want to know if the product is available in blue.
Nobody answers before 9am the next day. By 9am, they've already bought elsewhere — from someone who replied at 10:05pm.
This scenario repeats itself hundreds of times a day across Beninese SMEs. WhatsApp has become the real point of sale. But a point of sale that closes at night and depends on a single person to answer.
A chatbot is not the same thing as a keyword menu
Most "WhatsApp chatbot" offers sold today are decision trees: the customer types "1" for hours, "2" for prices, "3" to talk to a human.
That's not nothing. But it's not an intelligent chatbot — it's a 1990s phone switchboard, disguised as WhatsApp.
The problem with this model: it remembers nothing. The customer who already asked for the price yesterday starts the conversation from zero today. They type "1" again. They retype their name, again. They re-explain what they need, again.
What I build is different because it starts from a simple principle: a system that doesn't remember a customer doesn't automate anything — it just moves the problem somewhere else.
What a WhatsApp chatbot must do, at minimum
Three things, in this order:
Answer frequently asked questions, 24/7. Price, hours, availability, location. The baseline. The customer gets an instant answer, even at 10pm.
Qualify before handing off. The chatbot asks 2 to 3 questions (budget, specific need, timeline) before escalating the conversation to a human. The salesperson stops wasting time on window-shoppers — they only see prospects who've already answered the basic questions.
Hand off cleanly. As soon as the conversation goes beyond what the bot can handle — a complaint, a negotiation, an edge case — it transfers to a human with the context already summarized. The customer never starts from zero.
A chatbot that checks these three boxes saves real time. A chatbot that only does the first line — FAQs — is just a webpage bolted onto WhatsApp.
The real price, in Benin
The ranges circulating in the region for this kind of project run from 500,000 to 3,000,000 FCFA upfront, plus 50,000 to 100,000 FCFA a month in maintenance — a wide spread that hides the same confusion as with websites: "chatbot" doesn't describe a standard product.
What I charge for a chatbot that genuinely does the three things above (FAQ, qualification, handoff with context): 500,000 to 800,000 XOF, delivered in 3 to 5 weeks.
What moves the price within that range:
- The number of FAQs and scenarios to cover
- Connection to an existing product catalog (WooCommerce, Google Sheets, custom database)
- Appointment-booking integration (shared calendar, time slots)
- Language: French only, or French plus Fon/local languages
Beyond that — agents that execute actions (create an invoice, update stock, chase an unpaid bill) rather than just answering — that's a different project tier, not a chatbot. More on that below.
How it works, without the jargon
The customer writes on WhatsApp. The message arrives via the WhatsApp Business API. A system reads the message, searches its knowledge base (your FAQs, your catalog, your hours) for the answer, and replies — while keeping in memory everything said earlier in that exact conversation.
If the question falls outside the known scope, or if the customer shows a clear buying signal, the system hands the conversation over to a human — with a summary of what's already been said, not just "a customer is waiting for you."
The difference with a keyword chatbot: here, the knowledge base can evolve (new products, new prices) without rebuilding the whole decision tree. And per-customer memory lets you pick a conversation back up three days later without starting over.
The most common mistake: automating everything at once
The temptation, once you see what a chatbot can do, is to want it to handle everything — including complaints, price negotiations, edge cases.
It always backfires. A frustrated customer stuck in a bot's loop leaves for good — when a human might have saved the sale.
The rule I apply: the bot handles what's repetitive and low-stakes (FAQ, availability, hours). Anything involving money, a complaint, or an exception goes to a human immediately — not after three failed attempts by the bot.
What it costs you not to have this
Every message left unanswered overnight or during peak hours is a customer whose reflex is to message the next competitor on their list.
A salesperson manually answering "how much is it?" a hundred times a week isn't selling while they answer that — they're doing data entry.
The math isn't "how much does the chatbot cost" but "how many sales am I losing this week because I answer too late, or because my salesperson spends their time on questions a system could have handled alone."
And after the chatbot — business agents
A chatbot answers. An agent acts.
The natural next step, once the knowledge base exists and conversations are captured: agents that don't just answer anymore, but execute — create a quote, chase an unpaid invoice, update stock, log what they did in a reviewable journal.
That's a bigger project (full business intelligence), not an extra checkbox on the chatbot. But it's the natural direction once the first level is running.
A WhatsApp chatbot that forgets everything from one conversation to the next doesn't save anyone time — it just pushes the customer's frustration later into the day.
One that remembers, that qualifies, and that knows when to stay quiet and let a human speak — that's the one that actually changes how an SME runs.
Want this kind of automation for your business? Describe what you need — I'll reply within 24h with a timeline and budget estimate, no commitment.
→ Automating MECeF quotes and invoices — the same logic applied to invoicing → See all services — from local visibility to business agents → Detailed pricing by project type