
Dreamforce 2026 is a wrap and I am still buzzing.
Something felt different this year. The energy on the floor, the conversations in the demo pods, the dinners… customers weren’t asking if they should scale AI and agents. They already have. Now they’re asking how to do it without losing control of cost, governance, or quality. That’s the real shift.
And that’s exactly where Agent Fabric kept coming up, over and over. It wasn’t a pitch. It was the answer customers were already reaching for. The connective tissue that makes scaling agents actually work.
We showed up across San Francisco with one consistent message: the agentic enterprise isn’t a future state. It’s being built right now. And we’re here for it.
Here is a look back at everything that happened during Dreamforce 2026.
Agent Fabric keynote: from AI to ROI

Being on that stage Wednesday felt incredible. The speakers were energized, the audience was locked in, and the conversation in the room matched the moment. The question has shifted from “should we invest in AI” to “why aren’t we seeing returns yet.” That 94% stat landed hard, and shows that the overwhelming majority of enterprises are still are not seeing a strong return on their AI investments
Three things I keep coming back to.
- Governance unlocks speed. Capita cut handling time from 29 minutes to 14 while managing 350-plus agents on one control plane. Control and velocity aren’t opposites.
- Agents have to work where real work happens. ADT and Newell Brands showed what’s possible when you plug agents into the integrations you already have.
- Economics matter as much as capability. Token spend is outpacing returns at most companies. Visibility, hard controls, smart routing. That’s what separates AI that scales from AI that stalls.
We announced Agent Evals, a new capability that gives teams a scalable way to test agent behavior, auto-generating test scenarios, scoring outcomes against defined success criteria, and tying failures directly to remediation. All on the same platform as governance.
| “You can’t govern what you can’t see, and you can’t scale what you can’t govern”Tiina Stephens, Director of AI Engineering, Capita |
On sustainable cost, what struck me most in this chapter was how universal the pain was in the room. Token spend is moving faster than anyone projected and there is no clear line back to value. Real-time visibility, hard spending controls, and intelligent routing that matches model cost to task complexity were presented as the path to AI economics that actually scale.
The keynote closed with a new pricing and packaging model, a roadmap extending evals across APIs, MCPs, and LLMs, and a Golden Hoodie award for MuleSoft Ambassador Brad Ringer from AT&T. A well-earned moment to close out a big day.

MuleSoft sessions, now on-demand
The keynote was just one part of the MuleSoft presence at Dreamforce this year. Across the city, we ran dozens of breakout sessions, theater presentations, and workshops. Thousands of developers, IT leaders, and architects got hands-on time with Agent Fabric across every format.
Some of my favorite moments of the week came from our conversations with customers at some of our demos and sessions.
- MuleSoft Keynote: From AI to ROI with Agent Fabric
- Why Agent Fabric Is Your Trusted AI Control Plane
- Govern APIs, Agents, and LLMs with MuleSoft Omni Gateway
- Ship Fast, Stay Secure in the Agentic Era With Agent Fabric
- Demo: From AI to ROI with Agent Fabric
Missed out on one of our demos, or were unable to attend in person this year? We have everything on-demand at Salesforce+.
Beyond the breakouts

Dreamforce would not be Dreamforce without a few moments that had nothing to do with slides. Dreamfest brought Usher and Gwen Stefani to the stage for a night that the crowd will not forget. And during the day, the basketball activation at the IT & Security Zone Campground drew a crowd of its own.
To close things out, I want to say to every customer, partner, and community member who joined us this week: thank you. From the conversations at the Campground, the questions from the breakout sessions, the conversations at the Data Lab, all of it drives what we build next.




