Here’s my hot take: Salesforce just announced 30 new AI features for Slack, and I’m betting most of them will gather digital dust within six months. Not because they’re bad—but because they’re solving problems that bot builders like us already cracked years ago.
Salesforce rolled out this AI-heavy makeover in 2026, building on their January update that gave Slackbot agentic capabilities. Thirty features sounds impressive on a press release, but as someone who’s been building bots in Slack since before it was cool, I’m skeptical. The real question isn’t whether AI can live in Slack—it’s whether these features actually understand how work happens.
The Agentic Capabilities Hype
Let’s talk about what “agentic capabilities” actually means. In January, Salesforce updated Slackbot to act more autonomously—making decisions, taking actions, and handling tasks without constant human input. That’s genuinely useful. An agent that can pull data from your CRM, update tickets, and route requests? That’s the kind of automation that saves hours.
But here’s where I get nervous: 30 new features suggests a kitchen-sink approach. When you’re building bots, less is almost always more. I’ve seen too many enterprise AI tools collapse under their own weight because they tried to do everything. Users get overwhelmed, adoption tanks, and you’re left with an expensive chatbot that nobody uses.
What Bot Builders Actually Need
From my perspective in the trenches, the best Slack bots do three things well: they reduce context switching, they surface information at the right moment, and they get out of your way. That’s it. You don’t need 30 features—you need three features that work flawlessly.
The integration with company data and workflows that Salesforce is pushing? That’s table stakes now. We’ve been building custom integrations for years using Slack’s API. The difference is that we build exactly what each team needs, not a one-size-fits-all solution with 27 features they’ll never touch.
The Real Innovation Opportunity
What excites me about this announcement isn’t the feature count—it’s the signal that Slack is becoming a genuine AI platform. If Salesforce opens up these agentic capabilities to third-party developers, that changes everything. Imagine building custom agents that can tap into Slack’s native AI infrastructure without reinventing the wheel.
That’s where the real value lives. Not in pre-built features, but in giving bot builders the primitives to create exactly what their organizations need. I want access to the underlying agent framework, not 30 opinionated implementations of what Salesforce thinks work should look like.
The Adoption Challenge
Rolling out these features over the coming months gives Salesforce time to iterate, which is smart. But it also means teams will face a steady drip of new AI capabilities to learn, configure, and integrate into their workflows. Change management is hard enough with one major update—try doing it with 30.
I’ve watched teams struggle to adopt far simpler bot features because nobody took the time to train users or build the feature into existing processes. AI that lives “right in the flow of work” only works if people actually change their flow to accommodate it.
What I’m Watching For
As someone who builds bots for a living, I’m curious to see which of these 30 features actually stick. My money is on the ones that automate the most tedious parts of Slack: summarizing long threads, finding that one message from three weeks ago, and routing questions to the right person.
The features that try to be too clever—predicting what you need before you ask, or making decisions on your behalf—those will probably need serious tuning. AI confidence without accuracy is worse than no AI at all.
Salesforce is making a big bet that AI belongs at the center of workplace communication. They might be right. But as bot builders, we’ve learned that the best AI is the kind you barely notice—it just makes your work easier without demanding attention. Whether these 30 features can pull that off is the real test.
đź•’ Published: