By Jennifer Rivero & Shiva Goli
Every year, Dreamforce provides clues about where Salesforce is headed. Sometimes those clues come through a major product announcement. Other times, the bigger story is how the company is changing its approach to technology.
Over the last few years, Salesforce has introduced Data Cloud, Agentforce, and a growing number of AI capabilities. This year, we expect the focus to shift from introducing AI to showing how AI will change the way enterprise systems are built and used.
Four key developments stand out heading into Dreamforce 2026:
- Headless 360
- The Anthropic partnership: Claudeforce
- The Agent Marketplace
- Outcome-based pricing for Agentforce
Together, they point to a change in how Salesforce applications, AI, and business processes could work together.
Salesforce Is Going Headless
One of the most interesting concepts at Dreamforce this year is Headless 360. Historically, using Salesforce meant opening Salesforce. Employees logged in to access information and complete tasks. Customers used portals. Administrators configured the platform.
AI changes that model.
An employee may ask Microsoft Teams for customer information. An AI agent may retrieve information and complete part of a process. A customer may interact with a digital experience without ever knowing which Salesforce capabilities are working behind the scenes.
Headless 360 is important because it allows Salesforce functionality to be used without requiring every interaction to happen through a traditional Salesforce screen.
Data, business rules, workflows, skills, and actions can be made available to other applications and authorized AI agents.
That does not mean Salesforce interfaces are going away. It means they are becoming one way to interact with the platform rather than the only way.
Claudeforce connects AI to the Business
The Anthropic partnership is another sign that Salesforce is looking beyond a single AI experience. Organizations will likely use different AI models and interfaces depending on the work being done. The challenge will be giving those AI experiences access to the right business information without creating new security, data, or governance problems.
This is where Salesforce can provide significant value.
Claude may provide the conversational experience and reasoning. Salesforce can provide access to customer data, business processes, and actions within the right security and permission model.
The important question is not just: Which AI model is best?
It becomes: How do we connect AI to the information and processes needed to get work done?
We expect Dreamforce to spend more time addressing this challenge as AI becomes part of more everyday business activities.
Read the recent Claudeforce announcement here.
The Agent Marketplace
Another area we expect to develop is the Agent Marketplace. Organizations are experimenting with AI agents, but building every use case from scratch is expensive and difficult to scale. At the same time, companies cannot wait for Salesforce to create every capability they need for their specific industry.
A marketplace creates a middle ground.
Instead of starting with a complete agent, organizations could find and reuse smaller capabilities such as:
- Checking account information
- Retrieving contract details
- Looking up historical usage
- Initiating a service request
- Accessing market or pricing information
- Updating customer records
- Applying a business rule
These are not necessarily complete agents. They are capabilities that can be used as part of different processes.
For an energy company, the same capability might support customer service, broker interactions, an employee assistant, a portal, or a sales process.
The opportunity is to create reusable components that can support many different AI use cases instead of building every agent independently.
Paying for Results Instead of Access
Another major development to watch is how Salesforce continues to price Agentforce. Traditional software pricing is relatively straightforward. Customers pay for licenses, users, or access to a platform. AI introduces a different model.
Salesforce’s Pay per Resolution model for Agentforce Service points toward paying for completed work rather than simply paying for access to the technology.
Potential measures could include:
- Customer cases resolved
- Requests completed
- Transactions processed
- Tasks completed
- Workflows executed
This approach creates a closer relationship between technology costs and business results. However, it will also require companies to define what a successful outcome actually means.
If an agent starts a process and a person finishes it, was the work completed by the agent? How should the quality of an AI outcome be measured? How do companies compare the cost of an AI interaction with the cost of human work?
These questions will become increasingly important as AI takes on more responsibility for customer and employee processes.
So, what’s next in the world of Salesforce?
For us, Dreamforce 2026 is less about asking, “What new AI features will Salesforce announce?”
The more interesting question is: “How will AI change the way enterprise systems work together?”
Headless 360, the Anthropic partnership, the Agent Marketplace, and outcome-based pricing all provide part of the answer.
They suggest a future where AI is not another standalone application. Instead, AI becomes another way to access information, initiate processes, and get work done.
The technology may change. The interface may change. But the business processes and information behind the work still matter.
Employees will not always begin by opening a specific application. Customers will not always know which systems are being used to support their experience. AI agents may work across multiple platforms in the background.
But the business information and processes behind those interactions still need to be managed.
Salesforce is moving toward a model where those capabilities can be used in more places and by more types of users and applications.
What Should Organizations Be Thinking About?
Organizations do not need to wait for Dreamforce to start preparing for this shift.
A few questions worth considering are:
- What information and actions should AI be able to access?
- Where do our most important business rules live?
- Are our data and security models ready for AI to take action?
- Which AI use cases have measurable business results?
- What can we build once and use in multiple places?
- How will we measure the quality and cost of AI-enabled work?
The companies that get the most value from this next phase of AI will not necessarily be the first to adopt every new feature.
They will be the ones that understand how AI fits into their existing business processes and make deliberate decisions about where it can create real value.
