Google Dialogflow enables AI-powered virtual call center agents
Google has added beta features to its Dialogflow Enterprise service that allows companies to develop virtual call center agents. The virtual assistants can recognize sentiments based on artificial intelligence, among other things.
With Google Dialogflow it was already possible to create chatbots for customer support, now the possibility is added to set up virtual call center employees. They use speech recognition, predefined answers and features that use artificial intelligence to improve the answers or support real call center agents.
The service uses, among other things, Dialogflow Knowledge Connectors, with which the AI can analyze and interpret unstructured documents such as FAQs. The virtual call center agent can then draw information from this, in addition to automatic replies. When customers get on the phone with a virtual employee, they will be notified.
With Dialogflow Sentiment Analysis, the virtual operator can identify positive, negative and neutral sentiments among customers. The system uses the Cloud Natural Language API for this. Dialogflow already used this technique for chatbots. Furthermore, Google says it will improve the conversion of speech to text by enabling DeepMind’s Wavenet. The artificial intelligence can use information from previous conversations to recognize frequently asked questions or problems.
According to Google, the virtual call center agents are intended to complement real operators. The search giant calls it Contact Center AI and the intention is that customers first get a virtual assistant on the phone, who can answer questions. If that proves insufficient, the customer is transferred to a human employee. The artificial intelligence then remains active to support the helpdesk employee.
Google notes that Contact Center AI shares some underlying components with Duplex, but the underlying technology is different. Duplex is the virtual assistant for consumers, who can, for example, call a hairdresser to make an appointment.