TLDR
The Big Picture: Dreamforce 2026 was all about Salesforce positioning itself as the foundational data, systems, and business rules engine for AI agents.
Introducing AIforce: Salesforce’s live interface layer that allows any AI model to plug directly into Salesforce to access your company’s data, workflows, and business logic.
Claudeforce is in Open Beta: A major partnership with Anthropic that allows sellers to use Claude to pull live Sales Cloud data and perform tasks (like meeting prep or pipeline updates) natively, without switching tabs. A Salesforce admin must set it up.
Ironclad Security Promise: Salesforce CEO Marc Benioff stressed a strict “zero data retention” policy. AI models connected to Salesforce via official channels will absolutely not train on your proprietary company data.
SaaS Isn’t Dead: Countering the “Saaspocalypse” narrative, leaders (including NVIDIA’s Jensen Huang) emphasized that AI isn’t replacing SaaS software; it’s just layering on top of it.
Koa: Salesforce’s first CRM reasoning model that enables agents to reason through complex, multi-step tasks.
The Bottom Line: Because everyone has access to the same baseline AI models, your advantage lies in your organization’s unique proprietary knowledge, business logic, workflows, and operational experience. Structured data combined with explicit, well-defined process rules is what enables AI agents to function and deliver value.
Showmanship — the ability to dazzle and delight — has always been part of Dreamforce. From the star-studded stages and flashy demos to rocking out at Dreamfest, it’s the spark that captures people’s energy and attention.
This year, that showmanship took a cinematic turn. Matthew McConaughey opened the main keynote in a 3-minute trailer introducing AIforce as “the end of software that makes humans do all the work.”
From its very first moments, Salesforce made it clear that Dreamforce 2026 had just one storyline: Salesforce will be the trusted database and system layer for the Agentic Enterprise. This storyline was explored and expanded across multiple keynotes, but the two-hour kickoff with Marc Benioff, featuring cameos from Anthropic CEO Dario Amodei and Nvidia CEO Jensen Huang, revealed the major points and set the tone for the week.
AIforce
AIforce is Salesforce’s answer to multiple frontier AI models and evolving user behavior. People are spending more time in Slack and AI apps and less time in a browser, so every time a user has to log in to Salesforce, it breaks their flow instead of being part of it.
But the work that makes Salesforce so valuable — the data, workflows, business logic, semantics, permissions, security, governance — remains essential, regardless of the interface. With AIforce, any AI model can plug directly into Salesforce to access that foundation.
That’s Salesforce’s biggest pitch: it brings your business context to AI. And it’s a good one; at this point, it’s clear that an agent’s performance is only as strong as the data and systems supporting it.
AIforce’s main components are:
- Claudeforce: the Salesforce–Anthropic partnership
- Slackforce (Surfaces): which draws on Salesforce and other business context to create visualizations in Slack
- Agentforce Coworker: an AI agent built into Lightning
- Headless Toolkit: the set of APIs, MCPs, skills, plug-ins, and developer tools required to build and govern agents in Salesforce, regardless of interface.
With AIforce, Salesforce wants to bring flexibility and customization to any AI-centered UI. Onstage, Benioff made the case that Salesforce’s value extends beyond the browser, and this is the latest move to prove it.
Claudeforce for Everyone
Salesforce in Claude, the flagship plugin of the Claudeforce partnership, is now in open beta for everyone to use.
The tool plugs your Salesforce org into Claude and integrates a library of out-of-the-box skills (meeting prep, deal health reviews, pipeline updates) that sellers can use to prompt Claude instead of switching tabs. Claude will execute tasks by pulling live Sales Cloud data behind the scenes and returning the results in its native UI.
Patrick Stokes, President, Applications and Marketing at Salesforce, showed off a custom dashboard he built with Claude, complete with a custom Parker Harris avatar that shouted encouragement and advice from the screen. While most folks probably won’t be designing motivational avatars, it was a nifty demonstration of the creative flexibility now available to users.
How to turn on Salesforce in Claude
Unlike a personal Claude account, individual users can’t self-serve. Organizations have to opt in, and a Salesforce admin must set it up. Here are the steps to enable the beta:
- Request Access: A Salesforce administrator must request the plugin via AgentExchange.
- Configure Connection: The admin enables the Salesforce MCP server in the org and provides the OAuth client key and secret to the Claude organization admin.
- Enable in Claude: The Claude organization admin enters the OAuth credentials in Organization settings > Connectors and selects “Salesforce (Beta).”
- Assign Users: The plugin is assigned to specific groups within the Claude organization; sellers then sign in with their Salesforce credentials to access their data and the 37 prebuilt sales skills.
Organizations must also have the latest Sales Cloud Enterprise Edition to be eligible for the beta.
Salesforce’s Security Promise: Zero Data Retention
A business lives and dies by its proprietary knowledge. Salesforce has built its reputation on keeping that knowledge secure (they literally call their security framework and protocols the Trust Layer). That’s why Benioff made it crystal clear that Salesforce has a zero data retention policy that applies to every AI model that connects to the platform. AI applications cannot train on your data accessed via Salesforce.
It’s important to note that this applies ONLY to data accessed from Salesforce apps and platforms via headless configurations or authorized API connections. If your employees share proprietary or customer data housed in Salesforce via other ungoverned methods, that data may be subject to model training.
We’ll let Benioff have the last word on this one: “When you’re working in Salesforce apps, and when you’re working on our platforms, your data is your data. It does not go in the models. We’ve audited it, we’ve tested it, we’ve tried it. Regardless of what you’ve heard on media or from other vendors, let me assure you that Salesforce models are built with zero data retention, tested over and over again by our security teams.”
No SaaSpocalypse
We’ve all heard the media debating the “SaaSpocalypse” — how AI will rise to dominance out of SaaS’s ashes. But based on what we saw and heard at Dreamforce, that future is pretty unlikely.
Jensen Huang agreed with Benioff, calling the notion “nonsense” and saying “[AI] is going to be a layer on top of software.”
Salesforce reinforced that claim throughout Dreamforce, and many of its other announcements — from six pre-built agents joining Piper this year to Koa, a CRM reasoning model that helps agents complete complex, multi-step tasks — show how this SaaS giant is evolving to support an agentic future, no matter which AI model you choose.
The Salesforce Org Renaissance
Huang hailed AI as “a new industrial revolution,” saying, “Just as a few hundred years ago with electricity, we can power everything. With the internet we can find anything. And now with artificial intelligence as an infrastructure layer across the planet, we can know everything and do anything.”
But if AI is the new industrial revolution, Salesforce’s approach to connecting, contextualizing, and governing these models is a new renaissance, a rediscovery and return to foundational operating principles. All the work organizations have put into modernizing their systems, cleaning up data, and designing more efficient processes has become newly valuable.
Because when everyone has access to the same models and agents, what sets organizations apart comes down to proprietary knowledge, experience, and workflows. The challenge many face — and will continue to face — is translating that organizational richness into something AI agents can use.
Agents can’t work around missing information like a person would, so data gaps become gorges, while missing or vague instructions create expensive, token-hungry detours. Salesforce has made the interface customizable and interchangeable, but the work required to get value from the platform (and AI) remains the same.
While Salesforce is enabling business users to design their own agentic experiences and dashboards, best practices and governance must guide what they create and how. Our team at Coastal is riding that cutting edge, helping teams build reliable, secure AI solutions that start from their established foundation. If you’re curious about what AIforce means for your business, get in touch. We can have a grounded, realistic conversation about its capabilities away from Dreamforce’s dazzling stage lights.
