Chatbots
Answering quickly, without leaving anyone unanswered
A good conversational assistant handles what it can handle and passes the rest, with all the context, to the right person.
The problems we address
What brings you here
Requests arrive outside working hours
The customer waits until the next morning for a question whose answer takes three lines.
The WhatsApp channel cannot be managed by hand
Messages pile up on a phone, with no shared history and no follow-up.
Previous chatbots frustrated users
Rigid menus, dead-end loops, no way to reach a human.
Information changes and the bot stays out of date
Prices, opening hours, availability: without a connection to the real systems, the bot answers wrongly.
Deliverables
What you receive
- Assistant embedded in the website
- WhatsApp or other messaging assistant
- Support assistant connected to your knowledge base
- Assistant connected to an ERP or CRM to look up a real status
- Internal assistant for staff questions
- Manageable conversational knowledge base
- Handover to a human agent with the full history
Benefits
What it concretely changes
An immediate answer to common requests
Opening hours, availability, order tracking, procedure: handled in seconds.
No dead ends
When the assistant does not know, it offers a handover and passes on the context instead of going round in circles.
Up-to-date information
Answers about a status or availability are read from the source system, not from a frozen script.
A history you can use
Conversations feed the analysis of frequent requests and the improvement of your documentation.
Features
What we can do in this area
- Natural-language understanding in French and English
- Answers grounded in your documents, with sources
- Lookup of real data through APIs (order, case, availability)
- Collection of contact details and automatic creation of a request
- Escalation to a human with the context passed along
- Opening hours and different messages outside them
- Dashboard of conversations and unanswered questions
- Content administration with no technical involvement
Our approach
How we go about it
Analysis of real requests
We start from your existing exchanges to identify the twenty questions that make up most of the volume.
Writing the reference answers
Answers validated by your teams, versioned, and manageable afterwards without us.
Connecting the systems
Reading real statuses where it makes sense, read-only by default.
Internal testing
Your teams put the assistant through its paces before any public exposure.
Gradual opening
Going live, monitoring conversations, and continuously enriching the knowledge base.
Architecture
How it is built
The conversational assistant rests on the same architecture as our business agents, with a messaging channel at the front and an escalation path at the back.
Channels
Web widget, WhatsApp Business, internal messaging. One engine, several entry points.
Session
Conversation context, optional identification of the person, bounded history.
Knowledge
An index of approved content, filtered according to whether the person is external or internal.
Business lookups
Read-only calls to the ERP, the CRM or the booking platform.
Escalation
Creation of a ticket or a call-back request, with the full transcript.
Technologies
Relevant technologies
- TypeScript
- Node.js
- NestJS
- PostgreSQL
- pgvector
- Redis
- Socket.io
- WhatsApp Business API
- Docker
- Sentry
Industries covered
The industries where we apply it
Frequently asked questions
Frequently asked questions
Chatbots and conversational assistants
Can the assistant take an order or a payment?
What happens if the assistant does not understand?
Do we need a WhatsApp Business account?
Who updates the answers?
Africa Tech Services
A project around “Chatbots”?
A first thirty-minute conversation, with no commitment. We will tell you plainly whether we are the right partner — and if not, we will point you elsewhere.