Salesforce just dropped a new Slackbot, and this time it’s not the same old notification nagger. They rebuilt the thing from the ground up, turning what was essentially a glorified to-do list into an AI agent that can actually dig through your company’s data, draft documents, and take actions on your behalf.
It’s now generally available for Business+ and Enterprise+ customers, and honestly, it’s about time. The old Slackbot was cute for reminders—”Hey, don’t forget to add that person to the doc”—but it was never going to win any battles against Microsoft’s Copilot or Google’s Duet AI. This version is different.
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 a Porsche.” I appreciate the bluntness. He’s not wrong.
What actually changed
The original Slackbot ran on simple algorithms. It could suggest archiving a channel or remind you to tag someone, but that was about it. The new one is built around a large language model—specifically Anthropic’s Claude—with a robust search engine that connects to Salesforce records, Google Drive, calendar data, and years of Slack conversations.
Why Claude? Compliance. Slack’s commercial service has FedRAMP Moderate certification for U.S. federal government customers, and Harris said Anthropic was the only provider that could offer a compliant LLM when they started building. That exclusivity won’t last, though. He mentioned Gemini as a likely addition this year—”performance is great, cost is great”—and left the door open for OpenAI too.
Harris also echoed Marc Benioff’s line about LLMs becoming commodities. “I call them CPUs.” That’s a fair take. The model itself matters less than what you build around it.
No, they’re not training on your data
This is the part that matters for anyone worried about privacy. Harris was clear: Salesforce does not train any models on customer data. The reasoning makes sense—if you train a model on a confidential conversation, there’s no way to control who sees the answer. “Models don’t have any sort of security,” he said. So your sensitive Slack chats stay yours.
Internal numbers are impressive, but take them with a grain of salt
Salesforce tested this internally with all 80,000 employees. According to Slack’s CMO Ryan Gavin, it’s “the fastest adopted product in Salesforce history.” Two-thirds of employees tried it, 80% of those kept using it regularly, and satisfaction hit 96%—their highest ever for an AI feature.
Employees reported saving between two and 20 hours per week. That’s a wide range, and I’d love to see more granular data, but the organic adoption story is interesting. Within five days, employees created a shared Canvas called “The Most Stealable Slackbot Prompts” that now has over 250 prompts. Kate Crotty, a UX researcher there, found that 73% of adoption came from social sharing, not mandates from above. That’s how good tools spread.
What it actually does in practice
During a demo, product designer Amy Bauer showed Slackbot analyzing customer feedback from a pilot program, correlating it with a usage dashboard image, and surfacing insights. It can search across Salesforce, Google Drive, and years of Slack history to answer questions or draft summaries. For knowledge workers drowning in scattered enterprise data, this is exactly the kind of thing that might actually save time.
The bigger picture
This is Salesforce’s most aggressive move yet to position Slack as the hub for what they’re calling “agentic AI”—software agents that work alongside humans on complex tasks. It’s a direct shot at Microsoft and Google, both of whom have been pushing their own AI assistants hard. Microsoft has Copilot baked into Teams and the whole Office suite. Google has Duet AI across Workspace. Slack needed something that didn’t feel like a toy.
Whether this actually makes Slack the “front door to the agentic enterprise” as Harris claims remains to be seen. But it’s a solid step forward. The old Slackbot was a tricycle. This one might actually get you somewhere.
Comments (0)
Login Log in to comment.
Be the first to comment!