Local GenAI-boosted virtual agents have human-like conversations

By Ryan Falkenberg, 04 December 2023 – Originally published by IT Online.

Cape Town-based conversational automation specialist Clevva has recently boosted its virtual agents generative AI (GenAI) technology. This allows the virtual agents to have targeted sales, support and service conversations that not only feel human, but get the job done.

Virtual agents are very different to chatbots. They can hold conversations, just like a human, which means you feel like you are engaging an expert that will give you the help you need, not simply offer you generic answers.

“There is a lot of local interest in how GenAI can be used to improve contact centres and business. We have embraced Gen AI to enable powerful conversational automation that you can control. It’s a major shift,” says Clevva co-CEO Ryan Falkenberg.

The advancement was demonstrated live earlier this month at the national Business Process Enabling South Africa (BPESA) conference for the contact centre industry in Cape Town.

“When the industry saw our virtual agents in action, there was a stunned silence,” Falkenberg says. “There has been such an acceleration in digital technology development that the digital agents of just a few months ago are incomparable with the virtual agents we demonstrated. With GenAI, I believe we have finally cracked what we have been trying to do for over 10 years – automating rule-bound conversations in an adaptive, human-like way.”

Falkenberg says that GenAI, while capable of handling unstructured conversations, has been known to give non-compliant and inaccurate answers, or worse – to “hallucinate” and make up answers – if the prompts it is given are not specific enough.

This is because many people don’t know how to prompt GenAI to give targeted, specific answers. Clevva has overcome this by enabling its virtual agents to perform the role of ‘prompt engineers’, analysing each customer’s request before asking Gen AI to generate a response.

For example, a question like “What tourist attractions should I see when visiting Rome?” would traditionally deliver a generic list of top attractions in the city. If the virtual agent first asks the customer when they plan to visit Rome, what their personal interests are, how mobile they are and more, it becomes possible to better prompt the GenAI. Properly prompted, it can offer up a personalised itinerary that is tailored to that customer.

Another challenge with GenAI is that it still has a tendency to go “rogue”, and can’t be trusted to manage customer interactions unsupervised. In the contact centre space, GenAI is being deployed to support agents, but organisations still want human oversight.

“Clevva can create constraints around what questions a GenAI-enabled virtual agent can and can’t answer, pointing to specific data sets it may use and those that are out of scope,” Falkenberg says.

“By combining the almost-genius brain of GenAI with the rule-bound control given by virtual agents, businesses can use GenAI with increasing confidence. You no longer have to worry about customers asking your chatbot the ‘wrong’ questions and getting an inappropriate response. With a virtual agent they can control the conversation within your company’s accepted boundaries and ensure that every engagement is managed according to clear rules and processes,” Falkenberg says.

He adds that GenAI means organisations no longer have to teach their chatbot to understand what every customer is asking. With GenAI’s language understanding capabilities, virtual agents are immediately gifted with the power of language.

“GenAI understands sentiment and the virtual agent can then make the necessary adjustments instantly. Its ability to answer questions from data sets is incredible: where we used to have to pre-link to answers, now GenAI does it for us,” he says.

GenAI can also engage in multiple languages, which means virtual agents can become capable of serving customers in hundreds of languages.

“Our virtual agents are soon going to be able to converse via voice channels, so instead of typing you can simply talk to them. It’s like Siri or Alexa, except it can have relevant conversations, much like talking to a human expert,” he adds.

This means you will soon be having customer service conversations with a highly skilled virtual agent that offers a competent, human-like experience. And as a result, human agents can handle the ‘connecting’ conversations that talk more to the heart than the head.

View original article here.