It’s safe to say that chatbots are having a moment. There’s plenty of consumer interest — about 44 percent of respondents to a recent Capgemini survey said they’d order at a restaurant with the help of an automated assistant, while 44 percent said they’d use an assistant to undertake tasks like making bank transfers. Moreover, among people who already use chatbots regularly, they’re pleased with the convenience they afford — in that same Capgemini study, 87 percent said they’d had satisfactory money-transferring and food-ordering experiences.

Enterprises want to get in on the trend, unsurprisingly. And that’s where Nuance‘s new product comes in.

The Burlington, Massachusetts company today took the wraps off of Pathfinder, a virtual assistant and conversational analytics tool that taps terabytes of call center recordings to help dialog designers program virtual assistants. Unlike most chatbot development platforms, which require gathering and manually labeling data and extensively training natural language processing models, Pathfinder employs algorithms to suss out users’ intents, which it subsequently uses to group topically related conversations together.

Additionally, it uses the aforementioned call center data to build what Nuance calls a “conversation map,” or a decision tree of paths a conversation might take, that’s intended to uncover the most helpful answers to commonly asked questions and potential problem areas. Nuance says these maps are a good deal more efficient than the traditional approaches, which involve subject matter experts sitting down with conversation designers to script paths according to business processes.

With the insights and maps in hand, programmers can produce more natural and accurate scripts, Nuance claims, and in the process derive insights that lead to opportunities for strengthening customer relationships.

“Through Nuance Pathfinder, we are making the design and build of [virtual assistants] data-driven,” Thomas Hebner, head of product innovation at Nuance, said. “Pathfinder learns from the human agents — not just what they are talking about but how they guide customers through the transaction — reducing the time to make truly intelligent [assistants]. Because of Pathfinder, these kinds of highly intelligent [assistants] will be accessible to more companies than ever before.”

There’s a compliance element, too. By applying analytics in Pathfinder, clients can uncover key bits of knowledge that might otherwise get lost in the shuffle, such as whether call center agents are sticking to approved scripts. Hebner says that’s particularly valuable to industries such as finance and health care, where the demands for privacy compliance and conversational experiences can be inflexibly specific.

“Pathfinder helps large enterprises answer complex customer questions accurately and resolve problems quickly,” added Hebner. “For high stakes situations like these, Pathfinder is a game-changer that sets [us] apart and ahead of other AI players in the chatbot market.”

Nuance is no newcomer to AI. The firm’s natural language processing technology once powered Apple’s Siri, and in the not-too-distant past, the company teamed up with tech giants like Nvidia to power AI-driven radiology and conversational virtual assistant platforms.

Nuance in September 2018 announced Nuance Prediction Service (NPS), a new tool within its Customer Engagement Platform that enables brands to forecast client behavior and respond in an automated, contextualized fashion. And two years ago, it debuted Dragon Medical AI, an assistant designed to help doctors perform tasks like editing health records, reviewing patient charts, and filling prescriptions with voice.