Improving business processes and end-user service

Many public operators use the telephone as their primary means of communication with their clients. With a volume of 100,000 calls per year, handled by a team of advisors, this operator is facing several constraints such as limited availability of the service due to opening hours (9 am-5 pm), an insufficient number of calls picked up, and difficulty in managing peak activity. To improve caller satisfaction and gain in performance, this operator called on the Heka.ai teams to develop and test a conversational telephone agent (Voicebot), i.e., an automated relational device based on the artificial intelligence (AI) foundation.

• Implementation of a speech recognition solution (Speech to Text) and construction of a learning database.

• Develop an application solution capable of interacting with the telephony system and automating the responses given to callers. 

• Deployment of an API and a Web interface for the conversational voice agent's administration, management, and performance monitoring.

• Automatic processing of 25% of calls made by an operator (30k / year)

The press service of a public administration, which was in charge of annotating official photos taken during public events, called on our teams to build a facial recognition and automatic photo captioning application.

• Deep learning models detect faces in various photos and accurately identify the face.

• Automatic photo annotation GUI to verify the inference and enrich the reference database.

• Manual annotation is avoided on tens of thousands of photos.