Salesforce launches AIforce to bring its data and workflows into any AI interface

Salesforce has introduced AIforce, a new interface layer designed to make Salesforce data, workflows, business logic, permissions and actions available from AI environments without requiring users to work inside the traditional Salesforce interface.

The launch changes the role of Salesforce in an AI-driven workflow.

Instead of making employees open Salesforce, navigate records, build reports and manually trigger processes, AIforce allows an external AI interface to become the front end while Salesforce remains the governed system underneath it.

The first implementations include Claudeforce, Slackforce and Agentforce Coworker, connecting the Salesforce platform to Claude, Slack and Salesforce's own Lightning environment.

AIforce is built on Salesforce's Headless Toolkit and can expose platform capabilities through MCP servers, APIs, plug-ins and skills.

The architecture means an AI agent can potentially retrieve CRM information, reason across records, update Salesforce objects and trigger existing workflows while remaining subject to the user's Salesforce permissions and business rules.

........

COMPONENT	ROLE	WHERE THE USER WORKS	WHAT SALESFORCE PROVIDES
AIforce	Interface and orchestration layer	Any supported AI interface	Data, actions, permissions, workflows and business logic
Claudeforce	Salesforce inside Claude	Claude / Claude Code	CRM context, sales skills and Salesforce actions
Slackforce	Salesforce inside collaborative conversations	Slack	CRM context, live interfaces, records and workflows
Agentforce Coworker	AI teammate inside Salesforce	Lightning	Reasoning and actions across Salesforce
Data 360	Data and context layer	Underlying platform	Unified and federated enterprise data
Customer 360	Application and semantic layer	Underlying platform	Processes, business rules and permissions
Agentforce	Agent execution layer	Salesforce ecosystem	Specialized business agents

........

The product therefore separates where work is performed from where enterprise data and control remain.

··········

AIFORCE MAKES THE SALESFORCE INTERFACE OPTIONAL.

Salesforce is moving from being the screen employees use to becoming infrastructure that AI interfaces can call.

Traditional enterprise software assumes that users interact directly with the application.

A salesperson opens Salesforce, searches for an account, reviews its history, checks opportunities, updates a field and creates the next task.

AIforce allows that sequence to begin somewhere else.

A user working in Claude could ask for accounts with deteriorating pipeline conditions.

An agent could retrieve authorized Salesforce information, combine multiple records, explain the situation and potentially execute an allowed action.

The user therefore interacts with an AI interface rather than a CRM interface, while Salesforce continues to provide the underlying data model, permissions and transaction layer.

The same principle applies inside Slack, where a team discussing a customer could retrieve Salesforce context, create or modify records and trigger processes without moving into another application.

This architecture does not remove Salesforce from the workflow; it moves Salesforce behind the workflow.

··········

CLAUDEFORCE PUTS SALESFORCE DIRECTLY INSIDE CLAUDE.

A prebuilt MCP integration gives Claude access to governed Salesforce capabilities without requiring every company to assemble the connection manually.

Claudeforce extends Salesforce's partnership with Anthropic.

The initial implementation, Salesforce in Claude, packages Salesforce capabilities into a prebuilt MCP server that can be used from Claude.

Salesforce says the initial release includes 37 prebuilt sales skills covering tasks from prospecting to pipeline management and CRM hygiene.

A salesperson could ask Claude to inspect relevant Salesforce information, summarize an account, identify follow-up work or perform supported CRM actions without manually opening the corresponding Salesforce records.

The integration is not limited to conversational access: Salesforce is also providing a Salesforce Development plug-in for Claude Code with more than 40 development skills and access to a broader Salesforce skills library.

The structure creates two distinct Claude workflows around Salesforce: business users can interact with data and processes, while developers can use Claude Code against Salesforce development tooling and platform capabilities.

Salesforce says additional Claude capabilities are planned around Tableau analytics, service, marketing, commerce and industry-specific workflows, while Salesforce in Claude begins in beta.

··········

SLACKFORCE TURNS CRM DATA INTO A LIVE COLLABORATIVE INTERFACE.

Salesforce records no longer have to be consumed through a conventional dashboard before a team can work with them.

Slackforce applies the AIforce architecture to Slack.

Its most distinctive component is Slackforce Surfaces.

Instead of returning only a text answer in a channel, Salesforce can use information from Salesforce, Slack and other connected sources to construct a live interface that users can explore and act on together.

A team could ask to see the accounts at risk this quarter and why, then work from an interface combining CRM records with relevant conversations and operational context.

Slackforce also connects Salesforce to Slackbot, allowing the assistant to reason across conversational information in Slack and governed Salesforce context.

The product can therefore turn Slack from a communication layer into an operational interface over enterprise systems.

··········

SLACK CRM REDUCES THE NEED TO OPEN THE CRM APPLICATION.

Records can be created and modified from the conversation where the underlying business activity is already happening.

Salesforce is also introducing Slack CRM as part of Slackforce.

A large amount of customer information is generated in conversations before somebody records it in the CRM.

A sales call is discussed in Slack, a colleague mentions a new stakeholder, or a renewal issue appears in a channel, and somebody eventually has to translate that information into Salesforce.

Slack CRM shortens that gap by allowing users to create accounts, log call information and update records from Slack through natural-language interaction.

The practical value depends on how accurately AIforce translates informal conversation into structured CRM changes and how organizations configure review requirements for consequential updates.

··········

AGENTFORCE COWORKER KEEPS THE SAME MODEL INSIDE SALESFORCE.

AIforce also provides a native route for users who continue working in the Lightning interface.

Agentforce Coworker brings the same agentic model directly into Salesforce's Lightning environment.

Coworker can reason across account information, historical activity and other Salesforce context before surfacing insights or taking supported actions.

It can also call specialized Agentforce agents already created by the organization, allowing separate sales, service, renewal or approval agents to remain specialized while Coworker becomes the conversational layer that invokes them.

Salesforce says Coworker works with existing permissions and business rules rather than requiring companies to build a separate authorization structure.

··········

AIFORCE SITS ABOVE SALESFORCE'S EXISTING AGENT ARCHITECTURE.

The new layer does not replace Data 360, Customer 360 or Agentforce; it exposes them to more interfaces.

........

LAYER	SALESFORCE COMPONENT	PURPOSE
DATA	Data 360	Unifies and federates enterprise information
SEMANTICS + APPLICATION LOGIC	Customer 360	Defines business objects, rules, processes and permissions
AGENTS	Agentforce	Performs specialized autonomous or assisted work
INTERFACE	AIforce	Makes those capabilities accessible from AI environments

........

This four-layer structure explains why AIforce is different from simply connecting a chatbot to a database.

Database access provides information, while AIforce attempts to expose the meaning and operating rules attached to that information as well.

An opportunity record participates in permissions, validation rules, automation, approval processes, business logic and relationships with other Salesforce objects, and an AI agent needs those constraints if it is expected to do more than answer questions.

··········

MCP IS PART OF THE CONNECTIVITY LAYER, NOT THE ENTIRE PRODUCT.

Salesforce is combining Model Context Protocol with APIs, skills and plug-ins rather than betting on one integration mechanism.

AIforce runs on Salesforce's Headless Toolkit, which exposes Salesforce capabilities through MCP, APIs, plug-ins, skills and developer tools.

MCP gives compatible AI systems a standardized mechanism for discovering and calling tools and accessing context, but AIforce is broader than an MCP server.

Salesforce is packaging higher-level skills, permissions, business semantics and workflows around those connections, reducing the integration logic companies would otherwise have to build for every AI interface.

The difference is between a raw path — AI model → API/MCP → Salesforce records — and a governed path — AI interface → AIforce → permissions + semantics + workflows + agents + Salesforce data.

··········

EXISTING SALESFORCE PERMISSIONS REMAIN PART OF EVERY REQUEST.

The AI interface does not automatically gain unrestricted access to the underlying CRM.

Enterprise AI integrations become considerably more complicated once agents can modify operational systems.

A useful assistant might need access to customer information, while an autonomous agent may also need permission to update a record, launch a process or create another object, and those permissions are not equivalent.

AIforce is designed to route requests through existing Salesforce permissions and business rules, so the external AI interface remains constrained by the governed Salesforce environment underneath it.

Organizations will still need to evaluate which models are connected, what actions are exposed, how agent activity is logged and where human approval remains appropriate.

Moving the interface outside Salesforce makes governance more important, not less.

··········

THE BIG ARCHITECTURAL CHANGE IS THE SEPARATION OF SYSTEM OF RECORD AND SYSTEM OF INTERACTION.

Enterprise software may continue storing the data even when employees stop interacting directly with its screens.

AIforce illustrates a broader shift in enterprise software architecture.

Historically, the same vendor often controlled the database, the business logic and the interface.

AI agents allow those layers to separate: a company could continue using Salesforce as its CRM system of record while employees increasingly interact through Claude, Slack or another AI environment.

The resulting structure is: Salesforce as system of record and business-logic layer; an external model as reasoning layer; Claude, Slack or Salesforce itself as the user interface.

The application can therefore remain strategically important even when its traditional graphical interface becomes less central.

··········

DATA STUDIOS MAPPING: AIFORCE REDUCES APPLICATION SWITCHING WITHOUT REMOVING APPLICATION DEPENDENCY.

The visible software layer can shrink while dependence on the underlying platform remains unchanged or even increases.

A typical CRM task previously involved several user-controlled transitions: conversation → open Salesforce → locate account → inspect records → perform action → return to conversation.

With AIforce, the sequence becomes conversation → AI request → Salesforce context and logic → action → result inside conversation.

The number of interface transitions falls, but the number of underlying Salesforce dependencies does not.

The AI interface becomes useful precisely because Salesforce still contains the structured data, permissions, workflows and semantic relationships necessary to execute the request.

Poor CRM data remains poor CRM data, incorrect permissions remain incorrect permissions and broken workflows remain broken workflows.

AIforce can remove navigation friction, but it cannot automatically repair the operational structure underneath it.

··········

AIFORCE ALSO CHANGES WHAT ENTERPRISE SOFTWARE VENDORS COMPETE FOR.

Owning the user interface is no longer the only way to remain central to the workflow.

Generative AI creates a potential problem for SaaS vendors because users may spend more time inside general-purpose AI assistants and less time inside individual SaaS applications.

Salesforce's response is not to require the user to return to Salesforce, but to make Salesforce available wherever the user chooses to work.

Salesforce does not necessarily need to own every conversational interface if it continues to own the enterprise context and business operations those interfaces need.

The competition therefore moves from who owns the application screen? toward whose data model, semantics, workflows and agents are being called behind the screen?

··········

THE FIRST TEST WILL BE WHETHER AI AGENTS CAN PERFORM REAL CRM WORK RELIABLY.

Natural-language access is useful only if actions remain accurate, predictable and auditable.

Reading Salesforce information through Claude or Slack is the relatively simple part; the more consequential capability is taking action.

Updating an opportunity, changing an owner, starting a workflow, creating an account, triggering customer communication or invoking another agent all introduce a larger operational cost when the model misunderstands the request.

The value of AIforce will therefore depend on permission enforcement, tool reliability, auditability, structured outputs, workflow design and human approval controls, not only on model intelligence.

··········

SALESFORCE IS PREPARING FOR A WORLD WHERE USERS MAY NOT OPEN SALESFORCE.

AIforce keeps Salesforce underneath the work even when another AI product becomes the visible interface.

Claudeforce demonstrates the external-model route, Slackforce the collaborative route and Agentforce Coworker the native Salesforce route.

Together they show the same architecture from three different directions: the employee can choose the interface while Salesforce continues to supply the governed enterprise context underneath it.

If that model expands successfully, the Salesforce UI becomes one possible way to consume Salesforce rather than the mandatory destination for every CRM task.

The future value of Salesforce may depend increasingly on how much useful work other interfaces can safely perform through Salesforce, rather than how much time users spend navigating Salesforce itself.

··········

FOLLOW US FOR MORE.

DATA STUDIOS

datastudios.org