The days of having to spend your first day back after a weeklong vacation lost in Slack threads, trying to catch up on what your colleagues have been up to while you were away, are over. Or so says Slack. The reason is generative AI, natch. Now type after me: /giphy excited.
The workplace messaging app, which Salesforce acquired for $28 billion in 2020, has become the latest business productivity tool to integrate generative AI features. Starting on Wednesday, Slack is offering AI-generated summaries of conversation threads, and even entire channels, and will also have a new AI-powered search tool that will enable users to more easily find the information they need amidst an ocean of message traffic.
Noah Weiss, Slack’s chief product officer, said that the average Slack user will be able to save 97 minutes per week thanks to the new features.
The launch comes nearly one year after Salesforce first teased the plan to enhance Slack with AI capabilities. Denise Dresser, who became Slack’s third CEO since the Salesforce acquisition when she assumed the division’s helm in November, told Fortune she has been using the new AI-powered features to help her get up to speed on her new job. She has used it to summarize internal Slack channels. But she also has used it ahead of a key customer call to find specific points that a customer has been raising over the past few months when speaking with their sales reps.
“I think it just has an immediate impact in the way that you work and the way it drives even more productivity,” she said.
Slack is pricing the generative AI tools as a separate add-on to customers’ existing Slack subscriptions. But the company declined to reveal exactly how expensive the generative AI features will be, saying only that companies should talk to their Slack sales representatives about pricing.
Other generative AI features that are being integrated into business productivity tools have come with a hefty price tags that have made some customers blanche. Microsoft is charging an additional $30 per user per month to add its AI Copilot features to Microsoft 365, while Slack’s parent company, Salesforce, is charging customers $50 per user per month to access Sales GPT. These costs, as well as concerns about trustworthiness and security, have slowed the widespread adoption of generative AI, even as executives have clamored to experiment with the technology.
Dresser said that the new tools will help companies unlock the value of their own employees’ insights and expertise that don’t exist in structured databases, but instead are contained in Slack’s unstructured conversational threads.
Weiss said that when Slack was building the AI tools, they did encounter problems with “hallucination,” the term that describes when a generative AI models invents information. But he said the company had been able to engineer its search and summarization tools so that it only references information for which it can provide a citation to specific messages.
As a result, he said, the Slack summaries are accurate, but sometimes not fully comprehensive. “If we were asking to summarize a thread about the major features being discussed for a product launch, it will say x and y, and actually there also was z being discussed, but it wasn’t sure about z, so it won’t mention it,” he said.
Like many new generative AI search tools, the new Slack AI search gives the user a concise, capsule answer. But it also shows the user the threads that corroborate that information and allows the user to drill down into those messages to find out more. The search tool will also generate additional questions that a user might want to ask about the topic.
Users can choose to have automated summaries generated of the last seven days’ worth of conversation from a channel, or set a custom date range to summarize. The thread summary can reduce a long back-and-forth conversation to a simple paragraph.
Weiss, who has been at Slack for eight years, says Slack’s founders always dreamed of being able to use AI in this way. They realized early on that once a company had an archive of enough Slack messages, there would be value in mining those messages for insights. “Step 4 on our Master Plan from 2016 was ‘do magic AI stuff on top of the channel archive to give superpowers to knowledge workers,’” he says. At the time, though, the technology wasn’t there yet.
But after the debut of OpenAI’s ChatGPT in November 2022, it became clear that the technology was ready.
Relatively soon, Slack will also be integrated with Salesforce’s own Einstein Copilot generative AI product, allowing teams to use generative AI to create analytic dashboards, create Salesforce customer storefronts, and develop strategies for sales meetings, all from within Slack, Dresser says.