These days, you’d be hard-pressed to find a major hotel chain that doesn’t use social networking and media services to improve customer service. But many still handle this task with separate teams, ones that live apart from the call center or reservation desks. That can make for a disconnect.

Since June 2016, Hilton hlt has been taking steps to differentiate its strategy by adding artificial intelligence to the suite of digital concierge services it first began talking up two years ago. The idea is to convince more guests and travelers to use Hilton’s online reservation options on its website, rather than booking rooms using travel sites like Hotels.com.

Not only is that better for the hospitality company’s bottom line, it helps Hilton forge more ongoing relationships with the 34 million guests it serves annually, according to Mike Gathright, senior vice president of global reservations and customer care for Hilton. A successful experiment in the U.K. and Ireland convinced Hilton to roll out the service globally among its more than 3,000 reservation and guest assistance specialists, he says. Specifically, Hilton’s pilot of the system resulted in more online bookings through that region’s website.

“Hilton’s customer experience team is able to get in touch with the customer at a much earlier stage in the process—in the dreaming phase—than before,” Gathright tells Fortune. “As they’re browsing, there will be a chat box. As we sort of build the functionality over time, we are able to recognize over time what they are shopping for and use that contextual information to have an understanding of what type of search they are making. Then, the reservation specialists can use the given context to assist them in what they need specific to their interests.”

Get Data Sheet, Fortune’s technology newsletter.

While there are many different software applications behind Hilton’s digital concierge service, including IBM Watson ibm , the technology driving this new messaging application comes from [24]7 Inc., a 16-year-old software company based in Campbell, Calif. The firm has had a relatively low profile, but its list of around 450 customers is expanding with some pretty big names; in late October, Australian telecommunications firm Vodafone also announced a customer service chatbot built on the predictive [24]7 technology, joining the likes of Adobe adbe , American Express axp , Disney dis , Grainger, Merrill Lynch, and United Airlines ual .

Founder and CEO PV Kannan sold his previous startup to business software company Kana (which was, in turn, sold off to Verint). He’s been adding some pretty respectable talent to his executive team. The latest two hires, disclosed in early October are former Oracle orcl vice president Joe Eskew, who joined [24]7 as chief revenue officer, and former Veritas human resources chief George Haddad, who was named head of worldwide human resources. The company employs about 1,100 people as part of its corporate operations (excluding its contact center resources).

Despite the hype over the use of messaging as a customer service channel, which reached a fever pitch when Facebook repositioned Messenger as a resource for businesses, Kannan says only half of the most frequented online commerce sites use chat.

For now, at least, there’s an opportunity for companies to use predictive chat to differentiate their presence—especially those in the telecommunications, retail, and hospitality sectors, he says. “Usually when someone is hunting around on a website, they’re looking for something they need, but can’t find,” Kannan explains. “That’s when there’s a real opportunity for an agent to jump in, and make the conversation more person.”