Slack has announced a slew of new capabilities for its AI-enabled Slackbot, with greater support for connecting to third-party apps, taking meeting notes, and automating desktop tasks.
With expanded access to the model context protocol (MCP), Slackbot can now route user queries directly to Agentforce, as well as over 6,000 apps within the Salesforce ecosystem or any third-party tool an organization chooses.
Slackbot can sit in on and transcribe business meetings in apps such as Zoom, providing live responses to users based on enterprise data. Because Slackbot is connected to other business apps via MCP, it can automatically generate follow-up actions, populate CRM fields, and take other post-meeting actions.
Slackbot only listens when it’s asked to, at which point it draws on the audio feed of the user’s device. This means it works across different languages, Seaman added.
Organizations also now have access to reusable and schedulable skills for Slackbot, which leaders can share with employees to give them shortcuts to unlocking new features within the tool.
Seaman told assembled media that this could include a standardized approach for updating forecasts, which can be shared with all sales reps at the organization to embed it within their Slackbot.
In addition to its expanded capabilities within Slack itself, Slackbot will also gain desktop integration capabilities. This is the technology that allows it to listen to audio within meetings, which will also let it read a user’s screen in other apps such as for generating answers based on user emails.
Fluid Slackbot integration
With the expanded MCP support, any app within Slack can register its MCP server through Slack’s manifest, allowing Slackbot to call it when relevant.
“One of the most important things about this to me is that every single employee, with the explosion of these AI tools and agents, no longer has to think about which one to go to in the moment,” said Rob Seaman, EVP & GM of Slack,
“They can just turn to Slackbot and say, ‘hey, I have this particular sales strategy thing that I want to get accomplished and that will actually figure out what MCP server tool and agent it needs to call to get it done.”
Seaman added that this may result in an Agentforce agent being called, or even a ticket being sent to a human worker depending on the best solution to a user’s problem.
Users can also now request Slackbot perform ‘deep research’ on a question, prompting it to continue processing the request until the user’s demands have been met. Slack has additionally added memory to Slackbot, allowing it to gradually provide more relevant, personalized answers to user queries over time.
“So right now you can go in and ask Slackbot a question and say, ‘hey, you know what? I don’t actually care about the amount of time that you take to do this’,” Seaman explained.
“I want you to exhaustively research how this particular thing happened over the past, what learnings I might be able to get from it, and don’t just look at our Slack, look at our Google Drive, or our OneDrive that’s connected, and also look at the web as well.”
Secure AI within Slack
Whenever Slackbot searches Slack, takes an action, or accesses another system it does so as the user who is accessing it, to ensure privacy and security controls are applied.
“Slackbot cannot access anything that the user cannot,” Seaman said.
“So it can read public channels, it can read private channels the user is in, and it can read the user’s direct messages. When it goes off and searches Google Drive as an example, it’s searching as that user. That user has to OAuth in, so it does not have any super user access to any information, it’s not training across users, anything like that.”
These same rules apply when Slackbot acts on the user’s behalf in Google Workspace, such as creating a document, for example.
In response to customer requests, the company has also introduced ‘Confidential Channels’ within Slack, to which human users have access but Slackbot does not.
Slackbot success
Slack reports high user uptake of Slackbot, with Seaman clarifying that the tool is heading towards one million weekly users – a stat that would make it the fastest-adopted feature in Salesforce’s 27 year history. Some employees at customer organizations who already use Slackbot are saving 90 minutes per day, the firm added.
“Slackbot has become a vital resource for our teams,” said Peter Stoltz, VP, Head of CIO Office at reMarkable.
“Its natural language search is incredibly intuitive; you can ask a broad question, and it instantly pulls the right context from across DMs and channels to provide a clear timeline of events.
“Whether we are summarizing a busy week or confirming specific agreement terms, Slackbot consistently delivers the quick, accurate answers we need to stay aligned, saving our teams hours of manual work each week.”
Slackbot is currently available in the Business+ and Enterprise+ pricing tiers, with no associated cost tied to the number of tokens it consumes.
ITPro asked Slack if this would change to a cost-per-token model in the near future, as inference costs increased with features like deep research.
“We’re always evaluating how we package and deliver AI capabilities to reflect the value we provide customers,” said a Salesforce spokesperson in response.
“Today, powerful AI tools like Slackbot are included in existing Slack subscriptions, and we’ll continue to assess our approach as the technology evolves.”
FOLLOW US ON SOCIAL MEDIA
Follow ITPro on Google News and add us as a preferred source to keep tabs on all our latest news, analysis, views, and reviews.
You can also follow ITPro on LinkedIn, X, Facebook, and BlueSky.
