EM&AI: Building Conversational AI That Reads Human Sentiment and Intent

Written by Brandon Lee, originally published on KrASIA.

We've all had the experience of calling up customer service and navigating automated menus. Without a human representative on the other end of the line, the process can be frustrating—maybe the call ends up in a loop, or we never quite get the help that we need. This is the situation that Vietnamese startup EM&AI is attempting to remedy.

EM&AI has built a service based on conversational artificial intelligence that engages in active learning and humanlike communication. This involves the application of natural language processing (NLP), sentiment analytics, and text-to-speech and speech-to-text processing, according to the startup's chief marketing officer, Ho Minh Thi.

Over the past five years, EM&AI has refined its conversational AI to be used by financial institutions, insurance providers, and contact centers. The startup claims its AI is able to process human sentiment and generate responses that fit individual callers.

The startup achieves this by deploying natural language processing techniques to analyze and interpret the structure, meaning, tone, speed, and pitch of words of human speakers. EM&AI's algorithm then builds a sentiment model based on this information. Any information that isn't processed automatically is handled and labeled by the startup's data analytics team.

These capabilities also work in the other direction, where EM&AI's algorithm evaluates customer service representatives. It learns from these agents to create a vocabulary that fits specific contexts—banking, insurance, and other use cases—to the point where it is able to provide suggestions and feedback to customer service representatives when they are communicating with clients. This way, EM&AI functions like a junior-level employee who gradually gains experience and knowledge about their work.

The company says its conversational AI's capabilities are ahead of similar products offered by competitors, which solely offer speech-to-text and text-to-speech processing without taking into account customers' sentiments and intent during interactions. EM&AI claims its product is able to formulate a clearer understanding of customer responses, rather than merely follow a script and menu structure to route calls to human operators. It calls the process of shaping these conversations "hybrid chat."

On average, it costs USD 1 to USD 2.40 for a contact center to handle one call. Major enterprises handle roughly 3,750 calls each day, including inbound and outbound contact, meaning this carries a cost of USD 3,750 to USD 9,000. At the moment, EM&AI's services are offered at USD 1,000 per day. Plus, it streamlines interactions with customers and reduces the need to depend on human manpower.

To date, EM&AI has earned a total of USD 100,000 in revenue. It plans to bring in USD 2 million by the end of 2022 through monthly subscription fees from a growing pool of clients.

EM&AI was among the finalists of the Alibaba Cloud x KrASIA Global Startup Accelerator Vietnam-Indonesia Demo Day that was held on February 10.

Community

EM&AI: Building Conversational AI That Reads Human Sentiment and Intent

Read previous post:

Read next post:

Alibaba Cloud Community

You may also like

Comments

Alibaba Cloud Community

Related Products

IoT Solution

Offline Visual Intelligence Software Packages

CloudBox