Salesforce - Priscilla Hong

Dreamforce 2026 has wrapped up in San Francisco — three days, 1,600+ sessions, and a keynote where Salesforce did something almost no software company has the nerve to do: it declared its own interface optional. Co-founder Parker Harris put it on stage like this: "Why should you ever log into Salesforce again? Maybe you never will."

The pre-event announcements told us where Salesforce was heading. The keynote and Investor Day told us what actually shipped, what the numbers look like, and — most importantly for anyone building a career on this platform — which skills just became more valuable. Here are the conclusions that matter.

1. AIforce is the new headline — Salesforce becomes the context, not the screen

The single biggest launch of Dreamforce 2026 is AIforce, a "live interface layer" that moves Salesforce's data, business logic, permissions, and governance into whatever AI tool you already use. You describe the view you want; you get a working interface. No fixed tabs, no dashboard navigation.

It launched on three surfaces:

Every request runs on the asking user's existing permissions — an agent only sees what you can see — and the whole thing is built on Zero Data Retention: your data answers the question but is never stored or used to train external models. Under the hood sits Headless 360, exposing more than 60 MCP tools that any agent in Claude, ChatGPT, Cursor, or Gemini Enterprise can discover and call at runtime.

2. Koa: Salesforce's first CRM reasoning model

The NVIDIA partnership produced Koa — a reasoning model post-trained on NVIDIA's Nemotron family specifically for CRM logic: routing cases, evaluating opportunities, deciding next actions across structured business data. Two details are worth remembering for exams and interviews alike:

First, the training data is entirely synthetic — modeled on 27 years of enterprise workflow patterns across 14+ industries. No customer data was used. Second, it is CRM-specific by design — the argument being that general-purpose models guess, while business reasoning needs business context. Koa is in pilot with Formula 1 and UChicago Medicine, with US general availability slated for winter 2026.

3. Claudeforce is real, shipped, and already has users

The August partnership stopped being a press release this week. "Salesforce in Claude" ships as a prebuilt MCP server with 37 prebuilt sales skills (prospecting through pipeline hygiene), plus a separate Salesforce Development plugin for Claude Code with 40+ developer skills. Deloitte, GitLab, and Legora ran the early pilots. By Investor Day, Salesforce said up to 1,000 customers had signed up for the beta, with around 7,000 users already on Claudeforce.

4. Governance got its own product line

Hundreds of agents per company create a management problem, and Salesforce answered it with two launches:

Salesforce Guardian — an evolution of Shield and Trusted Services focused on agent identity (detecting when agents behave outside expected boundaries) and data security (classification and protection as agents touch sensitive records). Agent Fabric — a cross-platform registry where IT can discover, monitor, and govern agents built on Salesforce, Azure, AWS, and Google, tracking latency, error rates, and token costs in one view.

5. The proof points were unusually concrete

This year's keynote leaned on adoption numbers rather than vision slides:

6. The "SaaSpocalypse" was formally buried

For six months the market narrative was that AI agents would cannibalize SaaS. Benioff called that "crazy nonsense," and operating chief Miguel Milano went further: "I'm not going to use the word 'SaaSpocalypse' anymore. I hope you don't either." The rebuttal has a logic you should be able to articulate: AI models are probabilistic — they guess. Businesses run on deterministic systems where the number is the number. Value only appears when you ground the probabilistic layer in the deterministic one. NVIDIA's Jensen Huang made the same point from the hardware side: agents are an execution layer on top of enterprise systems, not a replacement for them.

7. Investor Day: $63 billion by FY2030

At the analyst session on September 16, Salesforce issued a FY2030 revenue target of over $63 billion — beating the $59.2 billion analysts had modeled — on top of FY27 guidance of $46.1–46.4 billion and a stated path into the $50 billion range next year. The bet behind the number: Gartner-size TAM math showing AI-related software spend growing at a 35% CAGR toward $2.5 trillion by 2030. The market bought it — the stock closed above $250, powered by the Anthropic investment's $2.6 billion paper gain and 27 consecutive years of growth.

8. What this changes for your certification path

Strip away the product names and every major announcement points at the same stack: trusted data underneath, reasoning and agents in the middle, governance on top. That is not a coincidence — it is exactly how the current exam domains are structured:

If you want to go deep on the data layer that everything at this Dreamforce was built on — ingestion, harmonization, identity resolution, and activations — the Data 360 Consultant practice tests cover these exact exam domains.

Bottom line: this was the Dreamforce where Salesforce stopped asking you to come to the software and started sending the software to wherever you already work. The interface changed. The skills that get you hired — data quality, grounding, governance — did not. If anything, this keynote made them the whole game.