Enterprise employees spend a lot of time jumping between systems to get even routine work done. Customer information might live in Salesforce, while the details needed to act on it sit somewhere else. That often means opening another application, tracking down the right information, and carrying that context from one system to the next.
Salesforce AIforce offers a different approach. Introduced at Dreamforce 2026, AIforce makes Salesforce data and business capabilities available in the AI tools and digital experiences employees are already using. Powered by the Salesforce Headless Toolkit, it allows authorized AI agents and applications to access Salesforce data and workflows or take approved actions without requiring every interaction to happen inside Salesforce.
AIforce sits within Salesforce’s broader Agentic Enterprise architecture. Data 360 provides unified business context, while Customer 360 brings the application logic and semantics that help agents understand how that information is used across Salesforce. Agentforce provides the agents themselves. AIforce extends that foundation into other AI environments.
For organizations that have spent years building processes, permissions, and business logic into Salesforce, that matters. AIforce provides another way to use those investments without recreating the underlying rules every time a new AI tool enters the enterprise.
The Salesforce Headless Toolkit makes Salesforce capabilities available through Model Context Protocol (MCP), APIs, command-line tools, Skills, plug-ins, and other developer tools.
MCP is an important part of that architecture. It provides a standardized way for compatible AI applications to discover the tools and context available to them. Rather than building a completely different connection for every AI experience, organizations can expose approved Salesforce capabilities in a way that MCP-compatible agents can understand and use.
Salesforce is launching AIforce with Claudeforce, Slackforce, and Agentforce Coworker, showing how that architecture can work across several different environments.
Claudeforce connects Salesforce directly with Claude through a prebuilt MCP server. This gives Claude access to Salesforce context without teams having to build the authentication and skill mapping themselves. It launches with 37 prebuilt sales skills, with support for additional Salesforce functions planned. Salesforce is also extending this approach to developers through its Claude Code plug-in and broader skills library.
Slackforce applies the same idea within Slack. Employees can bring live Salesforce context into conversations, update CRM records, and take other actions without opening Salesforce separately.
Agentforce Coworker takes the opposite approach by bringing the AI experience directly into Salesforce Lightning. Coworker can reason across Salesforce information and take action within the permissions and business rules already in place. It can also call specialized Agentforce agents an organization has already built and deployed.
The implementation differs across the three, but the architectural idea is the same. Salesforce remains responsible for the underlying data, semantics, permissions, business logic, and governed actions. AIforce gives agents and employees more ways to work with them.
Accessing Salesforce from other tools is useful on its own, but AIforce also changes how employees can work with the information Salesforce holds.
Think about a salesperson getting ready for a customer meeting. That person may need information spread across an account record, recent opportunities, service cases, activity history, and systems outside Salesforce.
An AI agent can work across more of that context at once. AIforce can reason across Salesforce data and logic, including information spread across multiple records, and pull in connected systems when additional context is needed. Rather than returning a list of records for an employee to sort through, the agent can use that context to surface information relevant to the request.
Data 360 can broaden that context further by bringing together harmonized and federated data, metadata, and memory. Customer 360 provides the Salesforce-specific semantics and application intelligence that help an agent understand what the data represents and what actions are available.
That becomes more important as agents move from answering questions to taking action. Knowing that a field contains a number is different from understanding what that number represents, how it relates to a business process, and what should happen next.
Connecting enterprise systems has not necessarily made them easier for employees to use. A workflow may be integrated behind the scenes while the person completing it still has to move between email, Salesforce, Slack, pricing systems, service applications, and other tools.
This constant application switching is sometimes referred to as the “toggle tax.” Employees lose time finding information and reorienting themselves as they move from one system to another.
AIforce creates the potential for more of that work to happen where the employee is already working. An agent with the appropriate access could retrieve Salesforce information, reason across the available context, and initiate an approved action without requiring the employee to manually navigate each underlying system.
AIforce also introduces more flexibility in how that information is presented. Salesforce describes the ability to create live experiences by describing what is needed rather than relying only on predetermined layouts and fields. An employee could generate a view around the task at hand, using Salesforce context to surface the information that matters for that particular workflow.
The traditional Salesforce experience still has an important role. What changes is that employees may not need to start there for every interaction with the platform.
Giving AI agents more ways to work with Salesforce raises an important question: what happens to the controls already built into the platform?
Salesforce designed AIforce so requests continue to run through existing permissions and business rules. An agent helping a salesperson update an opportunity, for example, should not gain broader access simply because the request started in Claude or Slack.
If the employee cannot view a record, the agent should not be able to retrieve it on that employee’s behalf. If a change requires approval or needs to satisfy an existing validation rule, initiating it through an agent does not remove that requirement.
That means organizations can continue building on the controls they already maintain in Salesforce rather than creating a separate permissions model for every AI tool employees use.
MCP adds another control point by allowing organizations to determine which tools and capabilities are available to compatible agents. Teams still need to decide what an agent should be able to discover, which actions it can invoke, and where human approval should remain part of the workflow.
Salesforce has also built AIforce around Zero Data Retention. According to Salesforce, business data used to answer a request is not retained by the model provider, an important consideration when Salesforce context is being used through an external AI tool.
Governance does not stop with access controls. Organizations will also need visibility into how agents interact with enterprise systems and what actions they take. As agents move from retrieving information to executing work, monitoring those interactions becomes increasingly important.
AIforce can connect to an organization’s existing Salesforce environment without requiring a new permissions model, data migration, or custom integration work for that connection. That can lower some of the technical barriers to making Salesforce available through new AI experiences.
The broader enterprise environment is where things get more complicated.
Salesforce may hold the customer record, while information needed to complete a workflow lives in an ERP, commerce platform, service application, or enterprise data environment. If an agent needs that information to complete its task, those systems and data still need to be available in a secure and usable way.
Data quality also becomes more visible. An employee working directly in Salesforce may recognize an outdated record or know that two fields are used inconsistently. An agent reasoning across that information does not automatically have the same institutional knowledge.
Business semantics matter for the same reason. Agents need more than access to data. They need reliable definitions for what that data means, how records and processes relate to one another, and which actions make sense in a given business context.
For Salesforce teams, this puts even more importance on the work already happening around data governance, integration, identity, and platform architecture. The quality of the AI experience will depend heavily on the enterprise foundation behind it.
That does not mean organizations need to expose every Salesforce capability to an agent from day one. A more practical starting point is a workflow where employees already spend significant time searching for information or moving between systems. From there, teams can identify the data the agent needs, define the actions it should be allowed to take, and determine where human involvement still makes sense.
AIforce gives Salesforce customers more ways to put the data, workflows, and business logic they have already built into Salesforce to work. The opportunity becomes much more meaningful when agents can also access the trusted enterprise context needed to complete a task.
Concord works across Salesforce, enterprise integration, data, and AI to help organizations build that foundation. Whether that means connecting Salesforce with other business systems, improving the quality and accessibility of enterprise data, or designing governed agent workflows, the focus is on making AI useful within the way the business actually operates.
Ready to explore where agent-driven workflows could fit into your organization? Connect with Concord to discuss your Salesforce environment and where AI could make the biggest impact.
No. Salesforce remains the underlying enterprise platform. AIforce makes approved Salesforce data, business logic, workflows, and actions available through additional AI interfaces while Salesforce continues to manage the underlying controls.
Agentforce provides Salesforce’s digital workforce of AI agents. AIforce is the interface layer that makes the broader Salesforce foundation available to AI experiences, including the data, semantics, business logic, permissions, and actions those agents need to work with Salesforce.
Model Context Protocol provides a standardized way for compatible AI applications to discover and use approved tools and context. Within AIforce, MCP can make Salesforce capabilities available to compatible agents without requiring every AI experience to use a completely different approach to accessing them.
Data 360 provides trusted, unified context by bringing together harmonized and federated data, metadata, and memory. That gives agents a broader understanding of the customer and business beyond an individual Salesforce record.
AIforce requests run through existing permissions and business rules. The agent sees information based on what the requesting user is authorized to access, while actions route through Salesforce and remain subject to the controls already in place.
Start with a specific workflow where AI could remove meaningful friction. From there, evaluate the Salesforce and non-Salesforce data the agent will need, whether that information can be trusted, and which actions the agent should be allowed to take. Organizations should also define where human approval remains necessary before expanding into more complex workflows.
Not sure on your next step? We'd love to hear about your business challenges. No pitch. No strings attached.
