Salesforce just rebuilt Slackbot from scratch — and it’s finally a real AI agent

Salesforce dropped a significant update today: Slackbot, that little notification bot you’ve probably ignored for years, has been completely rebuilt. It’s no longer just a reminder tool — it’s now a full AI agent that can search your company’s data, draft documents, and take actions for you.

Parker Harris, Salesforce co-founder and Slack’s CTO, put it bluntly: “The old Slackbot was, you know, a little tricycle, and the new Slackbot is like, you know, a Porsche.” I appreciate the honesty — the old Slackbot really did feel like a tricycle. It reminded you to add colleagues to docs and archived channels. That’s it.

The new version is built on entirely different architecture: a large language model paired with a robust search engine that can pull from Salesforce records, Google Drive, calendar data, and years of Slack conversations. It’s generally available now for Business+ and Enterprise+ customers.

Why Claude is under the hood — and who’s next

Slackbot runs on Anthropic’s Claude, and there’s a practical reason: Slack needed FedRAMP Moderate certification for U.S. government customers, and Harris said Anthropic was “the only provider that could give us a compliant LLM” when they started building. But that exclusivity won’t last.

“We are, this year, going to support additional providers,” Harris said. “We have a great relationship with Google. Gemini is incredible — performance is great, cost is great. So we’re going to use Gemini for some things.” OpenAI is also a possibility. This multi-model approach makes sense — no one wants to be locked into a single provider when the landscape shifts every six months.

Harris echoed Marc Benioff’s view that LLMs are becoming commodities: “I call them CPUs.” That’s a fair take. The real value isn’t the model — it’s how you connect it to enterprise data and workflows.

The data privacy angle that actually matters

On training data, Harris was crystal clear: Salesforce does not train any models on customer data. “Models don’t have any sort of security,” he explained. “If we trained it on some confidential conversation that you and I have, I don’t want Carolyn to know — 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 a refreshingly honest explanation of why enterprise AI should never be trained on customer conversations. It’s not just policy — it’s technically impossible to enforce access controls inside a model’s weights.

80,000 employees tested it — and they actually liked it

Salesforce ran this through all 80,000 employees internally. According to Ryan Gavin, Slack’s CMO, “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% — the highest for any AI feature Slack has shipped. Employees reported saving 2 to 20 hours per week.

The adoption was surprisingly organic. Within five days, employees created a Canvas called “The Most Stealable Slackbot Prompts” that now has over 250 prompts. Kate Crotty, a principal UX researcher, found that 73% of internal adoption came from social sharing, not top-down mandates. That’s the kind of organic traction most enterprise tools can only dream of.

During a demo, Amy Bauer, Slack’s product experience designer, showed Slackbot synthesizing customer feedback from a pilot program, pulling in an uploaded dashboard image, and correlating usage data — all in one conversation. It’s the kind of workflow that normally requires switching between five different tools.

This is Salesforce’s most aggressive move yet to position Slack at the center of “agentic AI” — where software agents work alongside humans. They’re competing directly with Microsoft’s Copilot and Google’s Gemini integration in Workspace. The question is whether Slack’s existing user base will embrace this transformation, or if it’s too late to change what Slackbot means to people.

I think the real test will come when non-Salesforce customers start using it. If the search and action capabilities are as good as the demo suggests, Slackbot could become the reason companies stick with Slack instead of migrating to Teams. But if it feels bolted on or unreliable, the tricycle-to-Porsche analogy will feel like marketing hype.

For now, it’s the most interesting enterprise AI launch I’ve seen this year — and I’m genuinely curious to see how it holds up under real-world pressure.