As a corporate communications platform, Slack has become the de facto storage repository for an organization's knowledge, but that information is difficult to retrieve using traditional search tools. Today, Slack is introducing several new features designed to make information more accessible, including new AI-powered search tools and the ability to summarize information in channels.
Noah Weiss, chief product officer at Slack, said the platform naturally collects company information in an informal and unstructured way. The problem is finding a way to uncover that trove of hidden knowledge. “The upshot of all this is that a new wave of generative AI capabilities now allows us to extract entirely new meaning and intelligence from all the analyzes created over the years.” [on our platform]'' Weiss told TechCrunch.
Last May, Slack announced at the Salesforce World Tour in New York City that it would incorporate generative AI into its platform. This was a generalized call to action with his creation of SlackGPT, a unique generative AI designed specifically for content on the Slack platform.
Today's announcement is about putting that into practice in a more concrete way. Weiss says being able to summarize a channel's content simply means employees don't have to catch up after the holidays or read long threads to get the gist of a conversation. With channel summaries, when you request a summary, Slack's AI model generates a summary of every topic discussed, with references, and shows how the model created each part of the summary. . According to Weiss, this was a key part of the feature's design.
“You can drill down into any area and see all the detailed context. That's why we want to be transparent, build trust, and make sure we're open to the public about our work so people can learn more if they want to. “I was seriously thinking about doing that,” he said.
Image credits: slack
The company allows users to ask questions in a natural way, similar to ChatGPT, but with Slack content rather than more typical internet content, so users can ask, “What is Project Gizmo?” You can ask questions like “. SlackGPT then again uses sourcing to provide answers, allowing people to see where the answer came from and whether it can be trusted.
Image credits: slack
Each answer includes a quality check that allows the user to decide whether the answer is good, bad, or neutral, allowing the model to learn the quality of the answer and allowing system engineers to see how well the model is performing. You can check if there are any.
He did not go into details about the underlying model, saying only that it is a mixture of large-scale language models. “What we found is that they all run differently and with different speeds and quality characteristics. He actually spent a lot of time fine-tuning the models to the data he had in Slack. We spent a lot of time and did a lot of work on the engineering side of Prompt as well.”
Slack AI with Search and Summarization is an add-on product to the Enterprise plan, so there is an additional cost on top of the regular license cost. Slack hasn't disclosed cost details, but the company says it's currently available in the U.S. and U.K., and is currently only available in English, but will be available in other languages in the near future.