The idea of this project was to create a fully-functional AI system that helps managers to train their communication and selling skills by practicing phone conversations with chat bots.
Up to 47%
of training success increase achieved.
of unique chat bots can be created.
RampEx Chat is a unique educational service for corporate workers, used for training them to successfully communicate with clients by phone and make sales. This service uses virtual reality to simulate business phone conversations, so that trainees, workers who deal with client service, and sales managers could develop and practice their communicative and selling skills with an assistance of specially educated voice chatbots. Talking to chatbots accelerates the training process and makes it more effective, because it involves AI and machine learning technologies. Data gathered from phone calls (communication with a chatbot) constantly grows and updates, which makes possible to create new and more realistic conversation scenarios.
The task was to create a unique application that allows practicing in talking to real clients by training communication skills in different scenarios with self-learning bots that reply to questions in real time. To make this project real and successful, we took a Chatter Bot library and modified it for creating bots with certain settings that respond to our client's needs. Our team had to modify the Machine Learning engine needed to create a bot that doesn't just correspond with a person, but communicates with the person vocally on the scenario. The project included a number of difficulties regarding the machine learning system, speech synthesis, and the client-side audio streams processing and editing. We needed to solve them all to make the bot's side of conversations most realistic. The Chatter Bot itself functions via text corresponding in the console. Since our goal was to achieve audio corresponding, we applied Web API Speech Recognition and Speech Synthesis in order to provide the system audio to get the text information and vice versa (the back-end system receives the text, analyses it, and sends the text to the client side where it is transforms into audio).
The Chatter Bot library was used for creating the functionality that allows to teach one bot a certain scenario so it can reply only to the questions in the scenario. However, if the client needs to create another chat bot, it will be identical to the first one, because they both take data from the same database. In terms of working on the project, we had to create many bots that differ from each other for separate scenarios. One scenario is one phone conversation with a bot trained to answer certain questions, thus, all bots must be trained differently and separately from each other. We edited and customized the Chatter Bot library to build good environment for creating unique training scopes for each bot. To solve this challenge, we created a different module in the database, which allows us to make bots trained for different scenarios.
Since this system was created for trainees' practice, an authorized manager should have a possibility to monitor the training process and its quality. To obtain proper monitoring the system required recording of each trainee-bot phone conversation. In the end of the conversation, the system builds a report with a sound record that allows a manager to place comments on it. The challenge our team faced here was to find a library that records voices of a real person (trainee) and a bot. We conducted research and saw that the vast majority of libraries cannot be used for ensuring required functionality, because they couldn't record and/or save the bot's sound on the server. We solved this challenge by using such software as a Speech Synthesis recorder, which recorded only the bot's sound, and Cranker JS to connect the trainee's audio piece with the bot's one.
The client received a fully-functional product that allows to train people to be professional B2B call-center managers. The functionality is presented as a user-person phone conversation with a chat bot. The system is applicable in different industries such as banking, insurance, sales, etc. The system is based on Artificial Intelligence, thus, the quality of bots improves constantly by the analysis and other processes made by authorized managers. In addition, the system has the resources to create an unlimited number of bots for different industries and communication training scenarios.