Slack released a feature May 3 that supercharges its search bar. Search a word or phrase, and in addition to the typical search results, Slack will begin to surface recommendations of people who talk about that topic often, and channels where more information on that topic might be found.
In other words, it’s going to help you find the right person to talk to instead of shouting questions into the void of public channels.
Noah Weiss, Slack’s head of search, learning and intelligence, says each internal Slack account will begin to analyze every public message it can access, looking at how often people mention keywords, whether they respond when asked a question and which channels are typically used for that kind of discussion. When people don’t talk about a topic for a period of time, their “expertise” will diminish in the eyes of the system.
“Channel names are pretty short. In some ways, they’re a lot like hashtags on Twitter, where they’re not necessarily self-evident,” Weiss says. “There might not be a channel about Tesla, but there’s a bunch of channels where Tesla is often talked about.”
Notably, the machine-learning model used to power the feature isn’t trained on specific keywords before it hits your organization, meaning it will adapt to the things written about most often for each Slack team. This means it might take a little while for the system to work, but, Weiss says—as many product managers have before him—the more data his product sees, the better it will get.
Internally, the feature was tested by deploying it to figure out who was the best person to ask about a code-named project—and then also to out the office Westworld-obsessive. The feature will begin to roll out to paid Slack teams with more than 50 people today.