Overview of Virtual Agents

EM&AI Virtual Agent is a virtual assistant (VA) building platform based on NLP (Natural Language Processing) technology. VA has the ability to have natural conversations, ensuring professional service

Concept of Virtual Agent (VA)

Conversational AI solution that can understand and automatically communicate with humans in natural language in the form of speech or text.

Virtual Agent technology

VA interacts with users similar to chatbots but is more intelligent & customizable thanks to Machine Learning (ML), natural language processing (NLP), context management... So the VA can remember and use the data collected to provide contextual responses that learn and get smarter over time.

For example: VA with the role to support ordering support will have the ability to save the conversation history and customer information. When the customer returns, the VA can identify the customer, their phone number, and delivery address without having to ask for information again.

Virtual Agent is used to assist humans in performing necessary & frequently repeated tasks. Thanks to that, businesses can optimally use human resources to solve more complex and specialized tasks.

An example of a business virtual assistant solution:

• Virtual assistant for loan consultation

• Virtual assistant confirms orders

• Virtual assistant for debt collection

• ...

Mechanism of action

Conversational AI is built on the EM&AI Virtual Agent platform with the following operating mechanism:

• Step 1: The customer sends a message to the virtual assistant.

• Step 2: The message is received and sent to the NLP (natural language processing) system for processing.

• Step 3: Based on the data on which the NLP system was trained, including utterance samples, intents, and entities, the virtual assistant will analyze the meaning of the customer's message content along with an estimated confidence score. This confidence score will determine the virtual assistant's response.

The virtual assistant provides pre-trained responses that correspond to highly confident detected intent and provides default responses for low confidence level to that customer’s intents (Example: Sorry, I don't understand, could you please clarify what you mean).