Slack Introduces AI Search and Summarization Tools

As a corporate communications platform, Slack has become a de facto storage repository for institutional knowledge, but accessing that information has proven difficult with traditional search engines. Slack today launched a couple of new features aimed at making that information more accessible, including a new AI-powered search engine and the ability to summarise content within chats.

Slack’s chief product officer, Noah Weiss, argues the platform organically collects company information in an informal and unstructured manner. The challenge is to discover a way to access that hidden mine of knowledge. “The punchline to all of this is that now this new wave of generative AI capabilities allows us to extract a whole new set of meaning and intelligence out of all that analysis that has been created for years [on our platform],” Weiss said in an interview with TechCrunch.

At the Salesforce World Tour in New York City last May, Slack announced the incorporation of generative AI into the platform. The invention of SlackGPT, its flavor of generative AI created exclusively for Slack platform content, was more of a broad call to action.

Today’s announcement focuses on putting that to work in more precise ways. According to Weiss, being able to summarise channel material allows staff to catch up after a break or simply avoid having to read a long thread to understand the core of the conversation. With channel summaries, you may request a summary, and Slack’s AI model will build a summary of all the subjects addressed, along with references to demonstrate how the model created each section of the summary, which Weiss says was an important element of the feature’s creation.

“You can drill down into any region, and we will provide you with full context. So we were thinking about transparency, developing trust, showing our work, and providing people the option to drill down to learn more if they wanted to, he said.

The company also allows users to ask questions in a natural manner, similar to ChatGPT, but it uses Slack material rather than more generalized internet content, thus a user may ask ‘What is Project Gizmo?’. Slack AI then provides an answer, complete with source, allowing individuals to understand where the answer originated from and whether they can trust it.

Each answer contains a quality check, in which users can indicate whether the answer was good, terrible, or neutral, allowing the model to learn about the quality of the responses and system engineers to assess how well the model is functioning.

He refused to go into detail about the underlying model, indicating only that it was a combination of huge language models. “We discovered that they all perform differently, with varying speeds and quality attributes. We spent a lot of time fine-tuning the models for the data we had in Slack, as well as conducting a lot of prompt engineering.”

Slack AI with search and summarization is an add-on product for enterprise plans, thus it will cost more than the standard license. Slack did not give pricing information, but it is now available in the United States and the United Kingdom in English only, with plans to expand into other languages shortly, according to the business.