AIforce - The New Salesforce Ecosystem

10/5/20264 min read

As Dreamforce 2026 came to a conclusion, the buzzword is AIforce. With all the advancements and developments in Salesforce happening at breakneck speed, it has become increasingly difficult for consultants and the like to stay updated. Einstein, BOTs, CDP, Data Cloud, Data 360, Agentic AI, Agentforce, MCP, Headless Toolkit, Claudeforce and most recently, AIforce have ended up confusing many consultants in the industry. How do these products fit in and which one should be used, and when? Will Agentforce continue to exist with the announcement of Claudeforce and AIforce? These are questions which need to be answered with clarity.

As the products and technology evolved, Salesforce has managed to navigate the AI bubble and stay relevant, much as it has done with the CRM market it helped define. Agentforce is Salesforce's platform for building and deploying AI agents that can automate and execute tasks across Sales, Service, Marketing and other Salesforce applications. The Einstein Trust Layer provides security and governance capabilities, the Atlas Reasoning Engine enables agents to reason and take informed decisions, and well-grounded business data provides the necessary context. Thus, Agentforce can help companies accelerate sales processes and enhance customer service. Using the new Agentforce Builder, Salesforce allows users to describe what they want their agents’ intended behavior using natural language. With the right instructions, actions, data and guardrails configured, an agent can be built, tested and eventually deployed. Salesforce distinguishes between enabling Data 360 and implementing Data 360. Its current Agentforce learning material says Data 360 must be provisioned and enabled for Agentforce use, but implementing it to connect, map, harmonize, and unify data is a separate step. Some of these features are currently in beta, with further details on availability, licensing, and pricing to follow.

Agentforce Agents

Agentforce agents can be analogous to robots or assistants that perform tasks based on how they are configured — through instructions and guardrails — and the data and context provided to them through grounding. An Agentforce Agent can comprise multiple subagents, each with its own instructions and actions. Associated channels define how agents interact with the outside world. Actions can be as simple as updating a field or summarizing a conversation and can be implemented using Apex, Flows and Prompt Builders. Agents are quite different from BOTs which relied on predefined rules and scripted responses whereas agents can reason using the available data, instructions and conversational context. Agents are also more dynamic in nature and support multiple languages.

Broadly, there are three ways to build agents. The recently released Agentforce builder accepts instructions in natural language, allowing an agent to be built using a more conversational approach. Agent Script can be leveraged by more tech-savvy developers for better control, while Agentforce DX is also available through Salesforce CLI & VS Code. Recent announcements also include several job-ready agents across various Salesforce products.

Testing any agent is paramount before deploying it. The Agentforce Builder offers Simulate and Live test modes. Agentforce Testing Center generates test cases and batch testing.

AIforce

AIforce has catapulted the Salesforce usage to new levels by reducing the need to log into the Salesforce User Interface. With AI increasingly becoming the new Salesforce UI, it encompasses Salesforce within other commonly used tools such as AI platforms like Claude and ChatGPT and collaboration platforms like Slack, using the Headless Toolkit. This live interface opens up a wide range of possibilities while retaining Salesforce's security and permission controls. Claudeforce, Slackforce and Agentforce Coworker, launched alongside AIforce, offer unique advantages. Data 360 provides the system of context.

Claudeforce

AI and CRM are stitched together now. Salesforce in Claude offers 37 pre-built skills ( skills act as pre-defined prompts, more details are in this link) which add value for reps, such as pipeline management and deal reviews. The development plugin works with IDEs and fast-tracks development. Claude is also available within Salesforce Agentforce, making the partnership more interesting and powerful.

Slackforce

Slack CRM, Slackbot and Slack Code combine to help users create and update records, pull data to create interactive views, along with AI-assisted development. Slackbots can process images ( ex: notes from a meeting) and take actions, help reps take informed decisions about clients and support a lot of other similar use cases.

Agentforce Coworker

Coworker runs within Salesforce, can be used outside of it using MCPs and can use the available context, agents and actions to assist users with tasks and workflows. Instead of users having to manually perform every step, the Coworker can determine the appropriate actions. It can also route requests to specialized and appropriate Agentforce agents that have already been built and deployed. It thus acts as the master space to get most of the work done instead of switching between apps- updating records, logging calls, resolving support cases, getting the summary of a deal in the pipeline among others are practical use cases. Using it is pretty easy now as it works as an enhanced version of the global search in Salesforce with the "Ask" button.

Headless Toolkit

As AIforce has virtually detached the Salesforce Platform from a fixed UI or browser, it uses the underlying Headless Toolkit to empower other platforms and access Salesforce capabilities by leveraging MCPs, APIs, CLIs and Skills. The Headless functionality relies mainly on metadata, APIs/MCPs, Skills and the Headless Experience Layer (HXL) which is the rendering engine that displays data in the front end. It is important to understand that the headless architecture follows a principle of " open access, not by-pass". Governance of the features being enabled and how third party systems which pull data from Salesforce behave are key factors to be considered.

MCP Servers, APIs, CLI & Skills

Model Context Protocol is an open standard for accessing Salesforce tools outside its UI. It's a plug-and-play mechanism to access, query and update data while respecting assigned Salesforce permissions. To give an example, a user using an AI tool like Claude or ChatGPT can request a report or ask to retrieve or update Salesforce data, and MCP provides the protocol through which the AI application can invoke the required Salesforce capability. Apart from the standard MCP servers available out of the box, custom ones can be created as well. MCP acts like a common adaptor between AI applications and external tools. MuleSoft MCP, Informatica MCP, Tableau MCP are a few other examples across the stack.

APIs and CLIs are relatively self-explanatory, while Skills are like command-based workflows which can be invoked when needed through AI tools. Agent Skills can be used to build applications, while Sandbox Skills can help manage environments.

Salesforce has stayed true to its number-one position and has been quick to introduce revolutionary and phenomenal additions to its ecosystem, as it usually does. The pricing needs further clarification at this stage as Salesforce moves towards bringing more of its AI features under the purview of Flex Credits.

Subscribe to my blogs here: