Freshworks Announces New AI Capabilities

Freshworks today announced new capabilities for its Freddy AI engine. By leveraging real-time behavioral data gathered by the Freshworks suite, the new tools are able to predict which customers are most likely to close, reformulate common FAQs via chat or voice, and proactively engage with users at risk of churning.

“This is a machine learning algorithm that looks at the deals in your pipeline and assigns a probability of close based on prior history and known characteristics of successful deals,” says David Thompson, chief marketing officer at Freshworks, of the ability to predict which customers are most likely to close. The algorithm can detect various signals that may not be readily apparent or available to sales reps. In the case of a free trial or freemium model for software, for instance, it could bring in signals such as whether a prospective customer is logging in, using, and populating the software with data—signals that indicate engagement and likelihood to close, Thompson says.

With regard to reformulating common FAQs via chat or voice, Freddy is now able to translate traditional static FAQs into conversational chat interfaces via text and voice chat. “Businesses spend a lot of time investing in writing FAQs, and customers spend a lot of time ignoring them,” Thompson says. “What we’ve discovered in our user testing is that FAQs are actually a lot more useful in different forms than just “go to the website and read the FAQ”; it’s much more useful either in a conversational format like a chatbot or a fully voice-enabled bot, which is something else we’re seeing gain traction among our customers.

“What that means is you write an FAQ and you formulate a question that you know or you believe your customers frequently ask and then you provide an answer, but the reality is if you’re in a chatbot or if you’re in a voice bot which our Freddy platform now supports, the question can come in many forms,” he continues. “Our algorithm can intuit different forms of the question that require similar answers. If the algorithm can’t answer a question, it will ask the administrator for the right answer, and the algorithm will become more intelligent as a result of that.”

As for catching users at risk of churning, Freshworks’ Freshdesk support software can now read user behavior on websites with an eye on catching signals such as rage clicking, confused cursor movement, and random scrolling. Freddy then alerts support teams, with the goal of catching support issues early. “This can be as simple as something like behavior on a website that indicates frustration, sometimes called ‘rage clicking,’ or if someone is having a problem on a website and they start clicking multiple times, or they do multiple abandons of a shopping cart, all of those types of signals can alert a support agent or a bot, depending on the situation, to intervene proactively with the customer,” Thompson says.