Dreamforce 2026 kicked off in San Francisco today, and Salesforce wasted little time making its direction clear.

AI is still at the center of the conversation, but the message feels different from the last few years. Salesforce is no longer talking only about putting AI inside CRM. It is increasingly talking about making the data, workflows, permissions, and business logic inside Salesforce available to AI wherever people choose to work.

That could be Claude. It could be Slack. It could still be the familiar Salesforce Lightning interface. The underlying idea is that users should not necessarily have to navigate through a traditional CRM interface to benefit from everything Salesforce knows about their customers and business.

That vision was at the center of Salesforce CEO Marc Benioff’s 10:00 a.m. PT opening keynote and several of the biggest Day 1 announcements.

AIforce Takes Center Stage

One of the biggest announcements from Day 1 was AIforce, which Salesforce describes as a new live interface layer for its platform.

The concept is important.

Salesforce already contains much more than customer records. For many organizations, it also contains workflows, automation, security rules, permissions, business logic, analytics, and connections to other systems.

Traditionally, accessing all of that meant opening Salesforce and navigating the application.

AIforce is designed to change that.

Salesforce says AIforce can make that same context and functionality available through AI interfaces, allowing users and agents to ask questions, update information, trigger workflows, and even create dynamic interfaces without necessarily navigating through traditional Salesforce screens.

Existing Salesforce permissions and business rules remain part of the equation. Salesforce is also emphasizing Zero Data Retention with external model providers as part of the architecture.

For Salesforce customers, this may ultimately be one of the more consequential shifts coming out of Dreamforce.

The question may eventually become less about which Salesforce screen someone should use and more about how Salesforce data and processes should be made available wherever that employee is already working.

Claudeforce Brings Salesforce Directly Into Claude

That strategy becomes much easier to understand when you look at Claudeforce, Salesforce’s expanded partnership with Anthropic.

The first major product coming from the partnership is Salesforce in Claude.

The new Claude plugin launches with 37 prebuilt sales skills designed around common seller activities, including account research, meeting preparation, pipeline review, prospecting, and CRM updates. Claude can work with Salesforce information using the user’s existing Salesforce permissions, and actions such as CRM updates can be routed back into Salesforce.

Think about what that could mean for a salesperson.

Instead of opening Salesforce, checking an opportunity, searching Slack, reviewing email, looking through previous interactions, and manually piecing together an account briefing, the seller could ask Claude to prepare them for the meeting.

Claude can reason across the available context, present what matters, help determine what requires attention, and then help with follow-up activities.

Anthropic announced today that Salesforce in Claude is now in beta.

And Salesforce is starting with Sales, but it does not intend to stop there. The company has already identified future expansion into areas including Service, Marketing, Commerce, and Tableau analytics.

Why Claudeforce Matters

There is a bigger story here than another Salesforce integration.

For years, Salesforce has been the place employees go to work with customer data.

Claudeforce starts to flip that relationship.

Salesforce can remain the trusted system containing the customer data, security model, workflows, and business rules, while Claude becomes one of the interfaces through which employees interact with it.

That distinction could become increasingly important as companies adopt multiple AI platforms.

Organizations may not need every employee to use the same AI interface. Instead, the challenge becomes making sure those interfaces are grounded in the same governed business data.

Salesforce Is Also Bringing the Same Idea to Slack

Claude was not the only interface highlighted on Day 1.

Slackforce applies a similar concept to Slack.

Salesforce says Slackforce will bring Salesforce context and actions directly into conversations. One new capability, Slackforce Surfaces, can assemble information from Salesforce, Slack, and connected systems into interactive experiences that teams can explore and act on together.

Slackbot is also becoming significantly more capable.

Salesforce describes scenarios where Slackbot could identify accounts that have gone quiet, examine Salesforce cases and related Slack conversations to understand what happened, create follow-up tasks, change ownership, and help draft customer communications.

Slack CRM takes that another step by allowing employees to create accounts, log notes, and update Salesforce records using prompts directly inside Slack.

The common thread between Slackforce and Claudeforce is becoming increasingly clear:

Salesforce wants its platform to power the work even when Salesforce itself is not the screen the employee is looking at.

Agentforce Coworker Keeps the Experience Inside Salesforce

Not every organization is going to want employees working primarily through Claude or Slack.

For those users, Salesforce is also introducing Agentforce Coworker as another AIforce experience within the Salesforce Lightning interface.

Coworker is designed to reason across Salesforce information and help users surface insights and take actions while remaining inside Salesforce and operating under the organization’s existing permissions and business rules.

Taken together, the three experiences illustrate Salesforce’s approach:

• Claudeforce: Salesforce intelligence and actions inside Claude
• Slackforce: Salesforce intelligence and actions inside Slack
• Agentforce Coworker: the same broader agentic concept inside Salesforce

The interface can change. The underlying Salesforce context remains.

Salesforce Introduces Koa, a CRM Reasoning Model

Another notable Day 1 announcement was Koa, Salesforce’s first CRM-focused reasoning model, developed with NVIDIA and based on the NVIDIA Nemotron model family.

Unlike a general-purpose language model, Koa is being designed specifically around CRM reasoning and complex, multistep business processes.

Salesforce says the model was trained using synthetic CRM scenarios rather than customer data. Salesforce also controls the model weights and runs Koa within its own infrastructure so customer information can remain within the Salesforce trust boundary during inference.

That is significant because enterprise AI increasingly requires more than generating a good response.

An AI system working inside CRM may need to understand that changing an opportunity stage can trigger other automation, that certain users should not have access to certain records, or that resolving a service case involves a specific sequence of actions.

The value of enterprise AI will increasingly depend on how well models understand that business context.

The Real Day 1 Story: Your CRM Is Becoming an AI Foundation

There were plenty of product names introduced on Day 1 of Dreamforce.

AIforce. Claudeforce. Slackforce. Agentforce Coworker. Koa.

It would be easy to treat each one as a separate announcement.

But there is a larger strategy connecting them.

Salesforce appears to be moving toward a world where CRM is not necessarily the application everyone has open all day. Instead, Salesforce becomes the trusted business foundation underneath a growing number of AI-driven experiences.

The CRM still matters enormously.

In fact, it may matter more.

Because AI needs context.

It needs accurate customer information. It needs clean account and opportunity data. It needs well-designed processes. It needs permissions. It needs business rules. And it needs clearly defined actions it can safely take.

An AI agent cannot compensate indefinitely for a poorly designed CRM.

What This Means for Revenue Operations Teams

For RevOps leaders, that may be the most important takeaway from Dreamforce Day 1.

The conversation about AI should not begin with:

“Which AI tool should we buy?”

It should begin with questions like:

Is our CRM data reliable enough for an AI system to reason over?

Are our sales processes actually represented in Salesforce?

Are our permissions and governance models ready for agents to take action?

Are critical customer interactions trapped in disconnected systems?

Which repetitive activities should an agent handle, and where should a person remain responsible for the decision?

Those questions become even more important if employees eventually interact with Salesforce through Claude, Slack, Agentforce, or interfaces that have not been introduced yet.

The interface may become flexible.

The underlying data and operating model cannot be.

Day 1 Sets the Direction for Dreamforce 2026

There is still much more to come from Dreamforce this week, including dedicated keynotes for Agentforce, AIforce, Data 360, and Slack.

But Day 1 has already established a clear theme.

Salesforce is betting that the next generation of CRM will not be defined by getting more people to click through more Salesforce screens.

It will be defined by making trusted Salesforce data, workflows, business logic, and actions available to people and AI agents wherever work happens.

Claudeforce may be one of the clearest examples yet of what that future could look like.

For organizations considering their own AI strategy, the takeaway is equally clear: preparing for that future starts long before deploying an agent.

It starts with getting the CRM foundation right.

Need help determining whether your Salesforce environment is ready for the next generation of AI and agentic workflows? Revenue Ops can help you evaluate your data, processes, automation, and Salesforce architecture before you make the next investment.