Salesforce Dreamforce 2026 Key Takeaways: Agentforce, Coworker and AIforce
This year’s Dreamforce theme, “Hey AI, Meet the #1 CRM,” struck a chord with me. It signals that Salesforce intends to embrace AI without abandoning its identity as the number one CRM company. Going in, I had been a little worried that Salesforce was drifting away from being a CRM company and trying to reinvent itself as an AI and data company. This theme put that worry to rest.
That said, AI was the undisputed focus of the event. Roughly 95% of sessions centered on AI-heavy concepts: Agentforce, Coworker, AIforce, headless architecture, MCP servers, and more. Here are my key takeaways from the conference.
AIforce: Salesforce, wherever you already work
AIforce is Salesforce’s bet on agentic, dynamic interfaces that let users go beyond standard Salesforce apps. Under the hood, a new headless architecture lets people tap into the Salesforce platform from whatever agentic tool they already live in.
Imagine using Claude, ChatGPT, or Copilot to directly query Salesforce data, or even to build dynamic dashboards and screens that let you interact with Salesforce from within your favorite AI tool.
WATCH OUT: Salesforce’s Claude integration is expected to arrive as a plugin; the beta version is free for now. What it will cost once that changes is still an open question.
Coworker: search grows into a conversation
Global Search has long been one of Salesforce’s most heavily used features, and Coworker is its natural next step, letting business users simply talk to Salesforce instead of hunting through menus. The demos alone were enough to win over a skeptical room.
WATCH OUT: Two things give me pause.
Agentforce has crossed into the mainstream
The shift from last year to this year was hard to miss. Back then, most of my Agentforce conversations were with skeptics. This year, I kept running into people who had already built and shipped their first agent.
This has now changed my advice to clients exploring AI agents: I now recommend Agentforce with real confidence, largely because of the trust Salesforce has built into the platform. Salesforce won’t let LLM vendors train on customer data, which means clients can put Agentforce to work on their data without worrying it will ever leave the Salesforce ecosystem.
Where I’d place my bets
Here’s how I’d size up the three AI capabilities Salesforce is pushing hardest right now:
| Rating | Capability | Guidance |
|---|---|---|
Agentforce | Not using it yet? Time to change that. Pick an agentic use case and take it to production — the tooling to track cost, validate results, and monitor agents is already mature enough to support it. | |
Coworker | Promising, but proceed carefully. Results need to be validated, and keeping a lid on costs takes real discipline. | |
AIforce | Still early days for production use. Treat it as a proof-of-concept exercise for now — get comfortable with what it can do before betting real workflows on it. |
The takeaway
Salesforce is going all-in on AI, but this year proved it isn’t willing to trade away the identity that built its business. Agentforce is ready for prime time. Coworker and AIforce have strong potential and are worth watching closely.
Consulting Partner
