Sebastian Thrun, one of the best known entrepreneurs in Silicon Valley, is taking on a new challenge that's a big shift from his work in autonomous transportation or online education:
He's working to automate sales chats.
Thrun, who founded Google's research lab X and its autonomous car project, education start-up Udacity, and electric aircraft company KittyHawk, is a co-founder and chairman of a stealthy enterprise company called Cresta AI. He's the elder statesman of the founding team, which includes Zayd Enam, Tim Shi, and Andre Esteva, who are all in their late 20s.
Cresta's artificial intelligence-powered system is designed to help automate mundane office jobs, allowing users to get work done more efficiently.
"If we train AI to take over boring repetitive tasks, we can free humanity and reach new heights," Thrun told CNBC.
Enam, Cresta's CEO, describes an unusual situation that he counts as a sign of the success.
A sales representative using Cresta's product recently told him that she got a sudden nosebleed while on the clock sending chat messages to customers, but was able to continue her work unimpeded. While ebbing the bleeding with one hand, she could click Cresta's conversational suggestions with the other.
The image of an employee working through the loss of vital fluids may seem bleak, but Enam says it illustrates the product's value. Just imagine all the other types of distractions you may need to work through or the mundane tasks that could be breezed through faster. The Cresta system finds patterns from successful experiences and serves employees chat recommendations based on all that data.
"We want to build artificial intelligence that makes people experts on the first day of their job," Enam said. "Ultimately we help sales agents make more money. They're having more effective conversations and they're having more of them."
The company currently works with three chat centers across the country.
Cresta got its start in Stanford's artificial intelligence research lab, where Thrun is an adjunct professor. Enam charts the genesis of the idea to a model he built using data about Udacity's human graders. Udacity students can rank the people who grade their work based on how helpful they are. The system Enam and Shi built could make recommendations to low-scoring graders based on the successes of high-scoring graders.
"We're trying to empower teams to operate like autonomous car fleets," Enam said. "When one person does something effectively, everyone else learns from their success, and when someone makes a mistake, everyone can learn from that too."
David Hornik, a partner at August Capital who has invested in enterprise software startups including Splunk, said that AI and machine learning are the next obvious step after big data to help companies work smarter.
"Anyone who's funded a company that has a call center knows that it's extraordinarily hard to keep those employees on the payroll — there's huge turnover," he said. "If there's an opportunity to remove the most menial and repetitive questions from the workload of those representatives, they'll have a better more engaging job and stay at the company longer."
It's still very early days for Cresta. The company has about 20 employees and recently raised a seed round of funding.