AI assistants are quickly making themselves at home across virtually every corner of the technology landscape. You’ll find them in enterprise platforms like Gmail and LinkedIn, and in more casual settings like Spotify, making everyday tasks smoother and more efficient.
Salesforce has been working extensively to enhance its AI features, and is now embedding its own assistive AI tool natively within Slack. These new features aim to take all of your company’s existing Slack knowledge – using previous conversations, recent huddles, and user profiles – to generate summaries, respond to queries, and filter important information that is most important to you.
For regular Slack users, this promises to be an exciting new feature that should really help streamline the way we read and comprehend conversations on the platform.
It also comes at an interesting time, given Salesforce recently updated their Slack API terms of service, ultimately preventing third parties from permanently storing their Slack data. They’ve created a competitive moat and are using their new data rules to their advantage.
To delve into this further, I spoke to Rob Seaman, Slack Chief Product Officer, to discuss all these new features and what users can expect from a new Slack experience.
What Are the New Features?
Last year, Salesforce first introduced Slack AI – the first step towards harnessing all of the informative unstructured conversations into actionable insights created by an AI assistant. This was then followed up with the Agentforce 2.0 update announcement, which introduced Agentforce onto the Slack platform for the first time.
Now, the future of Slack’s AI features is really starting to take shape. Today, Salesforce has launched a number of new features driven by AI that cover almost every corner of the Slack experience. Let’s look at the new key features that were introduced.
AI Writing Assistance in Canvas
Canvases are extremely useful when it comes to tracking action items and logging meeting notes, and now, Slack offers an AI writing assistant that can generate action items, refine content, and more with natural language prompts.
AI Message Explanations
If you’re not sure what something means, you can now hover over a message to get an AI-generated explanation, tailored to your team’s language and the context of the conversation.
AI-Generated Action Items
AI will pull out your most important to-dos and only nudges you when there’s real value, like a key update or something you shouldn’t miss.
Profile Summaries
You can now instantly see who someone is, what they do, and what they’ve been working on lately – perfect for getting up to speed without the back-and-forth.
Message Translations
When a message appears in another language, users can instantly translate it into their preferred language with a single click, making it easier to collaborate across global teams.
Enterprise Search
Slack’s unified search bar lets users find information across connected tools like Salesforce, Google Drive, Microsoft Teams, Confluence, and more, bringing together structured and unstructured data in one place.
Pricing and Availability
AI in Slack is now included across all paid Slack plans, marking a significant shift from last year’s model. Rather than offering AI as a standalone add-on, Salesforce has embedded AI features directly into the core Slack experience, with more advanced capabilities unlocking at higher tiers.
- Pro users get access to foundational AI features like thread and channel summaries.
- Business+ expands on this with features such as AI-powered search, translations, and workflow generation.
- Enterprise+ includes the full suite of Slack AI tools, including enterprise search, advanced task management, and enhanced security and compliance controls.
While not all features are available today, Slack has made several capabilities such as enterprise search, huddle meeting notes, and translations immediately accessible.
Additional tools such as AI message explanations, profile summaries, and writing assistance in canvas are expected to roll out soon.
In addition, Salesforce customers now have access to Slack’s free plan, providing a stepping stone to explore Slack before upgrading to unlock the full AI-powered experience.
Why Slack’s Context Makes AI Smarter
As you may already know, millions of informal conversations are taking place in Slack every day, and with that, information can often become disjointed or lost. The mission for AI in Slack, according to Rob, is to turn all of this unstructured context into actionable information that Slack’s AI can really act on.
“When we first introduced Slack AI, we began reimagining the Slack experience with AI built into its core,” Rob explained. “That’s critical, because the technology has finally reached a point where it can truly improve the way we work in Slack every day.
“The reason we’re able to make AI so powerful in Slack is because of its unique foundation: context. Slack holds your conversations, your files, your team’s priorities, representing both individual and company-wide work. It’s a deep well of institutional knowledge that makes AI, and AI agents, more effective. We see it evolving into an intelligent, agentic work operating system.
“And the news we’re announcing this week is all about advancing that vision by deeply integrating AI into every part of the core Slack experience.”
This brand new experience is all driven by one of Slack’s core principles, which prioritizes developing features that will help users work as quickly and efficiently as possible.
“We have a number of product principles, but the most important from a user perspective is this: don’t make me think,” Rob explained. “The software should get out of the way and simply help users get their work done.
“We want to apply AI in a way that reduces cognitive load so people can focus more on creative, high-value work. AI should handle the busywork automatically, working on the user’s behalf.”
How Does Slack Measure Response Success?
An AI assistant that reads through your data to give relevant responses sounds great. But in practicality, how does it filter through such an array of information correctly?
Most Slack channels are filled with unstructured, informal conversations between colleagues, and not every response is a reliable source of truth. So while Slack’s new AI tool has access to a wealth of data, the real challenge lies in filtering what’s actually useful from what’s not.
This is something Slack will be actively working on through a combination of their own internal teams and customer feedback.
“We have a model quality team that’s laser-focused on the accuracy and usefulness of AI responses. There’s a lot happening on the technical side, but the most straightforward part is the feedback we get directly from users.
“Every AI-generated response in Slack – whether it’s a summary, search result, explanation, or profile overview – gives users the option to give a thumbs up or down and provide feedback on what worked or didn’t.
“Participation rates are surprisingly high, some of the highest we’ve seen across any Slack feature, which gives us a strong signal. That feedback goes straight to our engineering team, along with the context, so we can quickly tune and improve the experience.
“It’s one of the clearest and most effective ways we continue to refine Slack AI over time.”
One of the biggest challenges this feature may face is the sheer volume of potentially irrelevant information. It will require Slack AI to grow very intuitively and read between the lines when it comes to delivering message explanations or search queries.
But as long as they’re transparent with their user base and accept feedback, this is likely something that will grow in substantial quality over time.
Are the New AI Features Still Effective for New Users?
As mentioned, Slack’s new AI features seem heavily dependent on having that additional context – months or years of conversational data – for it to perform well. But it raises questions as to how effective some of the features will be for companies just onboarding to the platform.
According to SQ Magazine, the platform added roughly eight million users to the platform in the last 12 months. There are many out there who are still getting to grips with the platform and won’t have that additional data compared to companies who have been committed for multiple years.
When I addressed this potential “cold start” problem with Rob, he was confident that the large language model (LLM) they’re using will have a decent level of knowledge to begin with, and will adapt quickly over time.
“The value of these features will definitely increase over time. In traditional machine learning – like old recommendation systems in retail – there used to be a real “cold start” problem, where you couldn’t generate useful results without a lot of prior data,” Rob said.
“But with LLMs, that issue is much less severe. Since they’re trained on a massive amount of global information, they start with a solid base of general knowledge, even before seeing any company-specific data.
“Where things improve over time is in understanding your specific company’s context. For example, if your company uses an acronym in a different way than the rest of the world, the AI won’t know that at first. But once people start using it in Slack conversations, the model can learn and adapt.
“So while there’s not a total cold start issue, it will take time – and active usage – for Slack AI to fully understand and reflect your organization’s unique language and context.”
It’d be fair to say that more recent users may be slightly on the back foot when it comes to using some of these new AI features – especially when it comes to message explanations, for example. But they will still have full access to some of the other key features in the meantime, while the AI gets up to speed.
It will basically mature as your Slack does – which is a pretty exciting reality.
What Slack’s New API Rules Really Mean
It was only last month when Salesforce announced their new API terms of service for Slack, which blocked third-party users from storing their messages permanently. This means that Slack’s data is locked tightly within the platform, making it harder for competitors to build products that rely on or enhance Slack data externally.
Salesforce officially framed the change as a proactive measure to strengthen customer data protection in an increasingly AI-driven landscape. But this, of course, raised some speculation within the ecosystem, and it seems we now know the key reasoning behind it.
With the rollout of Slack’s new native AI features, it’s clear that Salesforce is building tightly integrated tools that depend on deep, internal access to Slack’s conversational data. By limiting external access, Salesforce is tightening their data governance while creating a strategic foundation for its own AI ecosystem to flourish, without interference from third-party rivals.
Rob transparently explained that the aim of this was not necessarily to block users from using Slack data entirely, but to enforce more control over how the data is accessed and used.
“We could’ve communicated the Slack API changes more clearly. It’s a complex space with a lot of nuance,” Rob explained. “To be clear, we’re not trying to block third-party access to Slack data altogether. In fact, we want Slack to be a central hub for all kinds of AI, whether it’s built by Salesforce or by others. What we’re focused on is protecting customer data and enforcing our trust principles.”
He explained that going forward, Salesforce would rather third-party users go through the Slack App Marketplace so they can have a full view of who’s trying to do what.
“A big part of [this] is encouraging developers to go through our app marketplace, where we can review their apps, approve appropriate scopes, and ensure they follow Slack’s design and privacy guidelines.
“Some partners were bypassing that system, using APIs and permissions that weren’t reviewed, including access to private channels and direct messages. That creates risk, especially since Slack’s permissions can change over time.
“So the goal isn’t to shut people out. It’s to make sure third-party apps follow the right path. If they go through the marketplace, they’ll still have access to user-level data, including via our new real-time search API. We want a vibrant third-party ecosystem, just one that protects customer data and works as intended.”
So while the API changes initially sparked concern about Salesforce locking down Slack’s ecosystem, the broader picture is more nuanced. It’s less about shutting the door and more about tightening the framework to ensure that any AI built on top of Slack plays by clearly defined rules.
What the Future Looks Like for Slack AI
To round up our discussion, we asked Rob what the future of Slack AI really looks like.
The future of Slack, according to Rob, is the platform going beyond being just a communication tool and more of a central hub where users can trigger actions, monitor tasks, and interact with AI agents – almost like a smart dashboard for work.
“One way we think about it is as a kind of ‘command center’ or orchestration layer – that’s very agentic, right? Most users have Slack open for about ten hours a day and actively use it for around two of those,” Rob explained. So it’s a very natural launch point, not just for talking to colleagues or searching for information, but also for telling an agent, ‘Hey, I need this thing done,’ and having that agent take care of it.
“Or maybe the agent checks on a task and lets you know when it’s done. Over time, we see Slack becoming the place where you not only communicate, but where you coordinate and manage your work through intelligent, automated systems.”
It sounds as though we’re only just getting started with where Slack AI is heading, and the ambition is to scale it way beyond just a helpful assistant and one that, in future, will look to act on your behalf.
Final Thoughts
It was only natural that more AI capabilities would be added to Slack, but many of these new features are poised to fundamentally change how users interact with the platform day-to-day.
While some early teething issues are to be expected, this new approach feels like it’s here to stay and likely to grow more powerful over time.
