Dreamforce delivered many new and exciting announcements across the board this year – from admin and developer announcements to brand new Agentforce enhancements. But Slack also got its fair share of screen time during the three-day event, with plenty of new AI features being introduced that may change the way we interact with the communication app.
Slack has long been considered a collaboration layer within Salesforce, but the product is evolving quickly in the age of AI. Several AI tools are being developed to help users maximize their productivity, and could soon be the interface through which businesses can interact with their Salesforce CRM.
There have long been questions about Salesforce’s $27.7B purchase of Slack back in 2020. People have asked for more value from it, and with all of these new features being brought to the table, it seems Slack is slowly becoming more than just a chat add-on.
To delve into this deeper, I spoke with Laurence Fitch, Co-Founder of Slack-first Salesforce consultancy firm, Bryd, to discuss what to expect from Slack AI and how to overcome any potential roadblocks you may have today to get started.
What Are Slack AI’s Most Immediate Wins?
Slack AI is still very early stages of development. It’ll take time, feedback, and plenty of useful data before it matures.
But even now, customers are already enjoying quick wins, says Laurence, thanks to the Enterprise Search feature, which is now available across all the Enterprise+ plans.
“Some of the newest features that customers are most excited I would say across the board is enterprise search – the ability for Slack AI to search not just within Slack but within other data sources.
“So they’ve rolled out a list of pre-approved connectors – Google Drive, Jira, Microsoft Suite etc [and they’re] returning some great results. If I want to look for a deck, for example, for a financial services client proposal, the conversations live in Slack, but the content also in other places as well. Usually it lives in my Google Drive as well. So it’s great that it can return both of those using the Slack search tool bar, saving me switching between tabs.”
Drawing back to productivity, this saves users from having to search multiple sources – they can stay within Slack and still find all the information that they need quickly.
This is hugely beneficial, especially for those who work in RevOps or technical teams where they usually have to juggle multiple data silos and tools.
While a huge positive at this stage, Laurence admitted this is still something that Slack can build on: “People [also] want to be able to build custom connectors.”
Slack confirmed during Dreamforce week that this will be rolling out soon, which builds on an already big win for the platform’s current Enterprise Search model.
Laurence said: “Having that more open, where potentially partners or, if not customers, could build their own connectors into Slack enterprise search, I think would be a massive win or a bigger win on top of a pretty good win already.”
Channel Expert and In-Slack Agents
While Enterprise Search tackles productivity from a knowledge-management angle, Channel Expert takes aim at workflow efficiency, which is the next big area Slack AI is looking to transform.
“The ‘shiny’ one that caught people’s attention is Channel Expert,” Laurence explained. “It’s like a pre-release glimpse of what’s coming. It shows how Slack is becoming the place where you consume not just the Salesforce ecosystem but also your other SaaS apps and AI agents.”
One of the most significant challenges with AI adoption, he says, is that most tools live in another window. Users are bouncing between ChatGPT, Slack, and other systems, copying and pasting content back and forth between tabs.
OpenAI is looking to solve this, recently introducing GPT Atlas – enabling ChatGPT to live inside your browser while collecting extensive knowledge over time. Likewise, Channel Expert brings AI directly into the Slack channel where people are already working.
“The good thing about Channel Expert is the adoption and the setup are super-instant,” he said. “You drop it into a channel and there’s no complicated configuration. You can add relevant data sources, canvases, PDFs, Word documents etc., and it just works.”
According to Laurence, this feature can handle up to 80-90% of the routine questions that might otherwise hit an internal IT support queue. That means that human engineers can focus on higher-tier issues while teams get quicker responses to everyday issues.
Getting Your Slack Org AI-Ready
No AI feature can shine without solid foundations. This is an ongoing issue that we’ve seen already with Agentforce. Adoption rates have struggled due to technical debt preventing agentic AI from being as effective as it can be.
So for Slack, success also starts with clean, well-structured workspaces. Laurence explains how he has seen firsthand how “messy orgs” may hold back adoption of Slack AI.
“You can layer Slack AI on top of an organization,” he said, “but if all the conversations are happening in DMs or private channels, what can it find?”
He explains that most companies still might treat Slack as a chat tool rather than a collaborative knowledge system, which is something that may have to change in Slack’s AI era.
AI will rely on open, accessible conversations to learn and deliver accurate answers back to users. In order to meet in the middle, Laurence suggests that keeping channels open must happen to reap Slack AI’s benefits.
“The proper use of Slack (now) is an open-channel-first philosophy,” he explains. “Some customers even have an SOP (Standard Operating Procedure) template that says if you send a message as a DM or create a private channel, you have to justify it.”
By flipping the ratio – which means aiming for 80% public channels and 20% private, instead of the other way around – organizations can create an environment where Slack AI can actually find and surface useful insights for you and your team.
“Once you start communicating openly,” Laurence said, “Slack AI becomes smarter. It’s like a snowball rolling down a hill – the more it knows, the more valuable it becomes.”
Moreover, Slack’s new AI capabilities are most powerful when connected to Salesforce, even better when combined with an Agentforce agent deployed in Slack. But as Laurence points out, the same rule applies again: good data in, good insights out.
“Slack is a great UI for agents,’ he says. “But if the data it’s pulling from Salesforce isn’t correct, there’s no better way to ruin an AI experience. People make judgments about AI very quickly.”
This means companies must remain diligent about which fields their agents draw from, whether their Data 360 (formerly Data Cloud) setup is accurate or not, and if the AI can access the relevant sources.
“You’ll have game-theorized a hundred questions that people might ask, and they’ll ask the one-hundred-and-first that you weren’t expecting! So you need to make sure your backend setup can handle that.”
Slack may potentially now be the “front door” for AI, but Salesforce remains the engine driving it.
Data Migration and the Value of History
When Slack AI first launched, many teams may have seen it as a clean slate and a chance to start fresh. Before then, the goal was simply to improve collaboration, not to train AI. Historic chat logs from Microsoft Teams, Discord, or even WhatsApp were considered as clutter rather than valuable context.
There are multiple reasons for this – migrating old data into Slack isn’t easy, and is usually technical, time-consuming, and often more trouble than it’s worth.
“The two file types don’t really like each other,” Laurence explains. “Downloading that data, transforming it into something Slack can read, and uploading it can take six, eight, even twelve weeks.”
For many IT teams, starting fresh was cleaner and faster. Without AI-powered search or summarization, there wasn’t much value in pulling across years of messages that no one would ever read.
But Laurence stresses that Bryd is seeing this mindset change quickly. As Slack AI becomes more capable, historical data is valuable training fuel rather than baggage.
“Six to twelve months ago, when new customers came in – maybe from Teams or Discord – most said, ‘don’t worry about bringing the old data, we’ll just start again.’ But now, more and more are realizing all that history is worth something,” Laurence explains.
Drawing that data into Slack gives AI the context it needs from the first day, making search results smarter and summaries more accurate. According to Laurence, some customers start using Slack immediately while their old data imports in the background. “Someone might purchase Slack licenses and start using them straight away while they’re indexing their historic data behind the scenes. Then it floods in and gives everyone full context.”
As Slack continually evolves over the next few months, those who take time to prepare and properly migrate will enjoy the biggest payoff.
Final Thoughts
So, should you turn on Slack AI? For most teams, the answer is yes – but with a bit of preparation first.
As Laurence suggests, start by cleaning up your workspace, making channels public by default, and giving the AI as much useful context as you can over time.
Even in its early stages, Slack AI is already saving time through features like Enterprise Search and Channel Expert, and those benefits will only compound as the product matures.
