Slack, the prominent enterprise communications platform, has unveiled new features aimed at enhancing accessibility to institutional knowledge stored within its channels.
Over the years, Slack has unintentionally become a repository for corporate information in an informal and unstructured manner. To tackle the challenge of retrieving valuable insights from this vast pool of data, Slack has introduced a set of tools, including an AI-driven search function and the capability to summarize information within channels.
According to Noah Weiss, Chief Product Officer , the introduction of generative AI has opened up new possibilities for extracting meaningful intelligence from the extensive data accumulated on the platform. This move towards AI integration was first announced in May when they introduced SlackGPT, a tailored generative AI specifically designed for content within the Slack ecosystem.
Slack AI Features
The recent announcement focuses on applying this generative AI in more targeted ways. One key feature is the ability to summarize channel content, allowing employees to catch up swiftly after periods of absence or bypass lengthy threads.
By requesting a summary, users can leverage its AI model to generate a concise overview of all topics discussed, complete with references detailing how the model formed each part of the summary.
Transparency and trust-building were integral aspects considered during the feature’s design. Users can drill into specific areas to access detailed context, fostering a deeper understanding. Additionally, it enables users to pose natural language questions, akin to the functionality seen with ChatGPT. However, the responses are drawn from Slack content rather than general internet data. This approach ensures users can verify the source and reliability of the information provided.
Each response includes a quality check, allowing users to rate answers as good, bad, or neutral. This feedback loop aids in refining the model’s performance and assists system engineers in assessing its effectiveness.
While the specifics of the underlying model remain undisclosed, Weiss mentioned that it involves a mix of large language models. The team spent considerable time fine-tuning these models to align with the unique data within Slack.
It’s worth noting that Slack AI with search and summarization is an additional product for enterprise plans, incurring extra costs beyond standard license fees. While the company hasn’t provided specific pricing details, the feature is currently available in the U.S. and U.K., supporting English only at this point, with plans for additional language support in the near future.