Salesforce just dropped a completely rebuilt Slackbot, and honestly, the old one was a glorified sticky note. The new version is an AI agent that can search your company’s data, draft documents, and take actions for you. It’s a big swing in the agentic AI race against Microsoft and Google.
Parker Harris, Salesforce co-founder and Slack’s CTO, didn’t mince words: “The old Slackbot was a little tricycle, and the new Slackbot is like a Porsche.” I’ve been covering this space long enough to know when a company is overselling, but the internal numbers are hard to ignore.
What Changed Under the Hood
The original Slackbot did basic stuff — reminding you to tag someone, suggesting channel cleanups, pushing notifications. It was algorithmic and simple. The new one runs on an LLM with a robust search engine that pulls from Salesforce records, Google Drive, calendar data, and years of Slack conversations. It’s not a facelift; it’s a ground-up rebuild.
Salesforce kept the name because, as Harris put it, “People know what Slackbot is.” Fair. Brand recognition matters more than technical purity in enterprise software.
Why Anthropic’s Claude Got the Nod
Here’s the interesting part: Slackbot runs on Claude from Anthropic. Not because it’s the best model for every task, but because it was the only LLM that could meet Slack’s FedRAMP Moderate compliance requirements when they started building. That’s a real constraint for anyone selling to U.S. federal customers.
But exclusivity won’t last. Harris confirmed they’ll add Google’s Gemini this year, citing its performance and cost. OpenAI is also on the table. He echoed Marc Benioff’s line that LLMs are becoming commodities — “I call them CPUs.” I think that’s a bit dismissive of the engineering work behind these models, but from a product perspective, it makes sense to stay model-agnostic.
The Data Question Nobody Wants to Answer
On training data, Harris was refreshingly direct: Salesforce does not train any models on customer data. His reasoning was honest: “If I train it into the LLM, there is no way for me to say you get to see the answer, but Carolyn doesn’t.” That’s the right call. Once data is in the model weights, you can’t enforce per-user permissions. Keeping it out is the only safe approach for enterprise use.
The Internal Test That Actually Means Something
Salesforce rolled this out to all 80,000 employees for testing. According to Slack CMO Ryan Gavin, it’s “the fastest adopted product in Salesforce history.” Two-thirds of employees tried it, and 80% of those kept using it regularly. Internal satisfaction hit 96% — the highest for any AI feature Slack has shipped. Employees reported saving between 2 and 20 hours per week.
What caught my attention was how adoption spread. Kate Crotty, a principal UX researcher, found that 73% of internal adoption came from social sharing, not management mandates. Employees created a shared Canvas called “The Most Stealable Slackbot Prompts” that grew to over 250 prompts organically. That’s the kind of adoption you can’t buy.
What It Actually Does
During a demo, product experience designer Amy Bauer showed Slackbot synthesizing customer feedback from a pilot program, correlating it with a usage dashboard image, and generating an executive summary. It can search across Salesforce, Google Drive, and years of Slack history in natural language. No more digging through channels or hunting for that one spreadsheet.
The agentic part means it can take actions — updating records, scheduling follow-ups, triggering workflows — not just answer questions. That’s where the real productivity gains come from.
The Competitive Landscape
Microsoft has Copilot in Teams and Google has Duet AI in Workspace. Both are pushing hard. But Slack has an advantage: it was built for messaging first, and the new Slackbot is deeply integrated into that flow. Copilot feels bolted on sometimes. Slackbot feels native because it’s the same interface people already use.
That said, Slack’s market share is smaller than Teams. Enterprise adoption of Slack is strong in tech and creative industries, but Microsoft owns the broader office market. Salesforce is betting that agentic AI will be a differentiator, not just a feature catch-up.
Should You Care?
If you’re on Slack Business+ or Enterprise+, this is rolling out now. Try it. The internal numbers are unusually strong for enterprise AI, and the no-training-on-your-data policy is a genuine differentiator. If you’re on a free or standard plan, you’ll be waiting — and that’s probably intentional. Salesforce wants to drive upgrades.
The real test will come when external customers push it harder than internal employees. But for now, this is the most interesting enterprise AI launch I’ve seen this year. It’s not just another copilot. It’s actually built differently.
Comments (0)
Login Log in to comment.
Be the first to comment!