Botlhale, Africa's
Conversational AI

Let us help you engage with your customers in a language
they trust and understand.
Landing Page (circle 1)

Product Features


Rich Analytics

Helps you understand the user’s behavior, bot’s performance, usage pattern, and user behavior including sentiment analysis and major pain areas to drive action.


Cross Platform Integration

Creates an integrated and cohesive customer experience, across multiple channels, automating responses on whichever platform is most natural for the customer.


Conversational Bot
Builder Dashboard

Easy to use interface that allows the development of conversational AI. And with a library of prebuilt conversation templates for a variety of use cases across multiple industries, you don’t have to start from scratch.


Multilingual Capabilities

We offer all our services in 4 South African languages at the moment, namely: English, IsiZulu, IsiXhosa and Sepedi and we will soon be supporting Swahili, Afrikaans, Sesotho, Setswana and Shona.
We offer an end-to-end collection of services for creating conversation based digital interactions in African languages. Naledi is made up of seven Natural Language Processing (NLP) modules.

System Modules

Our system can be used and accessed independently or they can be grouped together to create more complex systems. These systems will also be offered in all South African languages. They use state of the art machine learning algorithms that are implemented in-house. The modules can be accessed through REST APIs or can be integrated on site as Microservices.
This module extracts the purpose of a user’s input. For example, if a user sent the following message to a banking bot, “Give Xolisani R50 from my savings”, this module is able to identify that the purpose of this message is to transfer money. In other words, this module would return “Money transfer”. This is useful in helping the banking bot respond appropriately or perform the relevant action.
The entity recognition module extracts a relevant term, object or value from a given input. For example, in the sentence “Give Xolisani R50 from my savings”, there are three entities relevant to a money transfer intent; Beneficiary, Amount and Account. This module would return:
Beneficiary: Xolisani
Amount: 50 rands
Account: savings
This module is responsible for controlling the state and flow of the conversation. It predicts the next best action based on the user’s current input and conversation history. For example, given the sentence “Give Xolisani R50 from my savings”, and assuming that Xolisani is not a known beneficiary. The next best action would be to ask the user what Xolisani’s account number is and who he banks with.

This module is part of the dialogue manager and is responsible for
customising dialogue states. It allows the dialogue manager to collect
multiple pieces of information through multiple dialogue turns,
before an action can be performed. It also handles conversation
detours, such as when a user goes off topic while collecting
information, the module keeps track of the information it has
collected and will loop back to complete the action when the user
goes back to the earlier dialogue state.

The TTS module takes in a text input and converts it to speech. The speech generated is supposed to be intelligible and it should sound as natural as possible. This module is useful in making computer generated voices. This can be used in dialogue systems like Siri or voice enabled GPS systems. This module returns an audio file from a given text input.
An ASR system is a system that converts speech to text, or simply, a system that allows machines to “hear” natural speech. This helps users interact with computer systems using speech. This module returns a transcription of a given speech audio file.
This module uses voice recognition to identify and validate a person. The module matches a speech phrase to an ID. This can be used in call centres, for example, to verify the caller’s identity.

Message us,
we would love to
hear from you.