Continuing its work with AI, Salesforce today announced it is rolling out native generative AI capabilities within Slack — or Slack AI.
Generally available starting today, the features empower users with built-in search and summary capabilities that make it easier for them to find the information they need in a matter of seconds. The same task could previously take several minutes to an hour, which affected user productivity.
“These new AI capabilities empower our customers to access the collective knowledge within Slack so they can work smarter, move faster, and spend their time on things that spark real innovation and growth. In the era of generative AI, Slack is the trusted, conversational platform that connects every part of a business to supercharge team productivity,” said Denise Dresser, who recently replaced Lidiane Jones as the CEO of Slack.
Slack is the home to all sorts of work conversations, starting from discussions on new deals and IT issues to marketing plans and HR policies. However, this collective intelligence – accrued across channels over the years – largely remains siloed, keeping users from making the most of it. If anyone needs some information, they have to do a lot of back and forth across channels and threads to find exactly what they need. This can take a lot of time, especially when a person is juggling multiple tools and tasks.
To make this easier, Salesforce is combining organizations’ collective intelligence with generative AI and launching three key features right within the Slack app – channel summaries, thread recaps and AI search.
With channel summaries and thread recaps, users don’t have to worry about missing out on conversations. All they have to do is click on a sparkle-shaped button and the models under the hood will automatically come into play, analyzing the messages in the channel/thread and providing a concise summary of everything discussed. Threads get a single summary while channels get a list, classified by most relevant topics.
This allows users to stay on top of conversations all the time, even when working in a different time zone or coming back from a long leave. Many can even use the feature with a fixed time period (like the last seven days or a month) to get a brief overview of everything that has been discussed and prepare for client meetings. According to Slack, this can save up to 30 minutes scrolling through messages and another hour of writing the summary.
Noah Weiss, the CPO at Slack, gave us an early look at the new features and confirmed that the underlying models are hyper-personalized to individual Slack instances, where they try to find what are the largest clusters of related conversation to define themes for channel summaries. However, what particularly stood out was the work done to ensure transparency. Both channel and thread summaries come with direct links that redirect the user to the source message, enabling them to learn more and validate whether the information is complete or not.
If the summary is not accurate, there’s also an option to rate it as ‘bad’.
“Because this is Slack, we have all the conversations in the product. This isn’t just a black box model; you can drill into any one of these summaries. So it shows you the citations at a message-by-message level for where it got the summary from,” Weiss said.
While channel and thread summaries analyze messages to produce a gist in the workflow of the product, AI search uses them to provide a Q&A-like experience from the search bar. This way, when a user asks a question about something, like what’s the leave policy, Slack will produce the exact answer and other related questions. Before this, searching on the platform produced tons of results linking to all the messages and channels where the search query was mentioned.
“It uses a deeper level of understanding of both what you’re asking and also what the conversations are about. Then, it matches those two together and synthesizes an answer that is actually combining multiple different conversations into one single answer. We have this on desktop but you can imagine that this will be even more life-changing on mobile. Getting the answer right before you head into a meeting is a total game changer,” Weiss explained.
That said, if the models do not find enough supporting evidence to provide the answer, they will just say so rather than hallucinating with incorrect information.
To deliver these capabilities, Slack took public models from third-party providers and fine-tuned them to run in its own virtual private cloud – making sure no data leaves the platform. However, Weiss did not name the models in use.
“We’re taking those models, running our environment, customizing it, and then applying it to help you leverage Slack data repository. We didn’t train any models on any Slack data or anything like that,” he said.
As of now, these AI capabilities only work in U.S. and U.K. English for subscribers of Slack Enterprise plans. However, the company said it is working to bring support for additional plans and languages.
It started working on these features about a year ago and has tested them with thousands of users from companies like SpotOn and Uber. On average, the tools helped these users save 97 minutes per week.
More importantly, Weiss confirmed that this is just the beginning of native AI features in Slack. In the coming months, the Salesforce-acquired communications platform will also launch a digest feature to help users keep up with major developments from channels that are not frequently used.
The idea with channel digest is you can move a large portion of your channels to be digest channels. Then, every day, you can get a one or two-line summary written by the AI of any really important conversation that happened kind,” Weiss explained. He also said that the company will bring Salesforce’s Einstein Copilot into Slack, allowing users to ask questions about all their customer data from Salesforce in the flow of working in Slack.
Continuing its work with AI, Salesforce today announced it is rolling out native generative AI capabilities within Slack — or Slack AI.
Generally available starting today, the features empower users with built-in search and summary capabilities that make it easier for them to find the information they need in a matter of seconds. The same task could previously take several minutes to an hour, which affected user productivity.
“These new AI capabilities empower our customers to access the collective knowledge within Slack so they can work smarter, move faster, and spend their time on things that spark real innovation and growth. In the era of generative AI, Slack is the trusted, conversational platform that connects every part of a business to supercharge team productivity,” said Denise Dresser, who recently replaced Lidiane Jones as the CEO of Slack.
Working smart with new Slack AI features
Slack is the home to all sorts of work conversations, starting from discussions on new deals and IT issues to marketing plans and HR policies. However, this collective intelligence – accrued across channels over the years – largely remains siloed, keeping users from making the most of it. If anyone needs some information, they have to do a lot of back and forth across channels and threads to find exactly what they need. This can take a lot of time, especially when a person is juggling multiple tools and tasks.
VB Event
The AI Impact Tour – NYC
We’ll be in New York on February 29 in partnership with Microsoft to discuss how to balance risks and rewards of AI applications. Request an invite to the exclusive event below.
To make this easier, Salesforce is combining organizations’ collective intelligence with generative AI and launching three key features right within the Slack app – channel summaries, thread recaps and AI search.
With channel summaries and thread recaps, users don’t have to worry about missing out on conversations. All they have to do is click on a sparkle-shaped button and the models under the hood will automatically come into play, analyzing the messages in the channel/thread and providing a concise summary of everything discussed. Threads get a single summary while channels get a list, classified by most relevant topics.
This allows users to stay on top of conversations all the time, even when working in a different time zone or coming back from a long leave. Many can even use the feature with a fixed time period (like the last seven days or a month) to get a brief overview of everything that has been discussed and prepare for client meetings. According to Slack, this can save up to 30 minutes scrolling through messages and another hour of writing the summary.
Noah Weiss, the CPO at Slack, gave us an early look at the new features and confirmed that the underlying models are hyper-personalized to individual Slack instances, where they try to find what are the largest clusters of related conversation to define themes for channel summaries. However, what particularly stood out was the work done to ensure transparency. Both channel and thread summaries come with direct links that redirect the user to the source message, enabling them to learn more and validate whether the information is complete or not.
If the summary is not accurate, there’s also an option to rate it as ‘bad’.
“Because this is Slack, we have all the conversations in the product. This isn’t just a black box model; you can drill into any one of these summaries. So it shows you the citations at a message-by-message level for where it got the summary from,” Weiss said.
While channel and thread summaries analyze messages to produce a gist in the workflow of the product, AI search uses them to provide a Q&A-like experience from the search bar. This way, when a user asks a question about something, like what’s the leave policy, Slack will produce the exact answer and other related questions. Before this, searching on the platform produced tons of results linking to all the messages and channels where the search query was mentioned.
“It uses a deeper level of understanding of both what you’re asking and also what the conversations are about. Then, it matches those two together and synthesizes an answer that is actually combining multiple different conversations into one single answer. We have this on desktop but you can imagine that this will be even more life-changing on mobile. Getting the answer right before you head into a meeting is a total game changer,” Weiss explained.
That said, if the models do not find enough supporting evidence to provide the answer, they will just say so rather than hallucinating with incorrect information.
Public models fine-tuned but language support limited
To deliver these capabilities, Slack took public models from third-party providers and fine-tuned them to run in its own virtual private cloud – making sure no data leaves the platform. However, Weiss did not name the models in use.
“We’re taking those models, running our environment, customizing it, and then applying it to help you leverage Slack data repository. We didn’t train any models on any Slack data or anything like that,” he said.
As of now, these AI capabilities only work in U.S. and U.K. English for subscribers of Slack Enterprise plans. However, the company said it is working to bring support for additional plans and languages.
It started working on these features about a year ago and has tested them with thousands of users from companies like SpotOn and Uber. On average, the tools helped these users save 97 minutes per week.
More importantly, Weiss confirmed that this is just the beginning of native AI features in Slack. In the coming months, the Salesforce-acquired communications platform will also launch a digest feature to help users keep up with major developments from channels that are not frequently used.
The idea with channel digest is you can move a large portion of your channels to be digest channels. Then, every day, you can get a one or two-line summary written by the AI of any really important conversation that happened kind,” Weiss explained. He also said that the company will bring Salesforce’s Einstein Copilot into Slack, allowing users to ask questions about all their customer data from Salesforce in the flow of working in Slack.
VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.
Author: Shubham Sharma
Source: Venturebeat
Reviewed By: Editorial Team
