What is Salesforce Agentforce? The Enterprise Agent Platform, Defined

Salesforce Agentforce is Salesforce's enterprise agent platform that runs autonomous AI agents inside the Salesforce platform, reasons over Data Cloud, and takes action on Salesforce records.

Salesforce Agentforce is Salesforce's enterprise agent platform, launched in 2024 as the successor to Einstein Copilot. It runs autonomous AI agents inside the Salesforce platform, reasons over Data Cloud and per-cloud records, and takes action on Salesforce records.

What Agentforce actually is

The pieces Agentforce is assembled from

Agents can be customer-facing, in service triage and sales prospect research, or internal, in HR helpdesk, IT triage and finance approvals.

Source: Agentforce per-conversation pricing and Thinklytics AI and automation practice, 2026.

Agentforce is the assembly of three pieces:

• The Atlas Reasoning Engine, Agentforce's orchestration layer that handles multi-step planning, tool calling, and grounding against the Salesforce data layer.
• The Agent Builder, the low-code authoring surface for defining agents, their tools, their guardrails, and their conversation flows.
• The agent runtime, hosted inside Salesforce's Hyperforce infrastructure, with billing per conversation rather than per message or per user.

Agentforce agents can be customer-facing (service triage, sales prospect research) or internal (HR helpdesk, IT triage, finance approvals). The platform supports both supervised (require human approval before action) and autonomous (act without per-step approval) configurations, with the supervision shape configurable per agent and per tool.

What people confuse it with

Einstein Copilot and Agentforce are not the same product

• Einstein Copilot. The assistant. The 2023 to 2024 brand for Salesforce's generative AI assistant.
• Agentforce. The agent platform. Launched in 2024. Adds autonomous multi-step action, agent-to-agent communication and the Atlas engine.

Agentforce is also not OpenAI Operator. Operator is a general-purpose browser-using agent. Agentforce is purpose-built for Salesforce-record-based workflows and reasons over Salesforce metadata natively.

Source: Thinklytics AI and automation practice, 2026.

• "Agentforce is just Einstein Copilot with a new name." Closer to true that it is the evolution. Agentforce adds autonomous multi-step action, agent-to-agent communication, and the Atlas engine. Einstein Copilot was the assistant; Agentforce is the agent platform.
• "Agentforce is Salesforce's version of OpenAI Operator." Wrong. Operator is OpenAI's general-purpose browser-using agent. Agentforce is purpose-built for Salesforce-record-based workflows and reasons over Salesforce metadata natively.
• "We can run Agentforce without Data Cloud." Mechanically possible. Operationally, the agent's reasoning quality degrades sharply without the unified profile Data Cloud provides.

When Agentforce matters

Where Agentforce fits, and where it does not

• Salesforce is the system of record for the workflow. The records the agent has to read and write already live there.
• Sales Cloud, Service Cloud or Marketing Cloud is the user-facing surface. The agent interacts with the screen the user already works in.
• Data Cloud is running or planned in the next 12 months. Without it the agent reads per-cloud records and the cross-entity context is missing.
• The workflow lives outside Salesforce. Microsoft-stack productivity, code generation and infrastructure operations belong elsewhere.
• No Salesforce platform investment. There is no record layer for the agent to act on.
• The agent has to reason over data Salesforce cannot reach. Reasoning quality is capped by the sources the platform can see.

We map the 6 to 12 workflows worth agent-ifying, score them against the Salesforce data foundation, then sequence over 6 to 12 months.

Source: Thinklytics AI and automation practice, 2026.

Agentforce matters when:

• The organization runs Salesforce as the system of record for the workflow.
• Sales Cloud, Service Cloud, or Marketing Cloud is the user-facing surface the agent has to interact with.
• Data Cloud is already running or planned in the next 12 months.

When Agentforce does not help

Agentforce does not help when:

• The workflow lives outside Salesforce (Microsoft-stack productivity, code generation, infrastructure operations).
• The organization has no Salesforce platform investment.
• The required agent has to reason over data sources Salesforce cannot reach.

How Thinklytics works on Agentforce

We scope Agentforce with workflow-first methodology. Map the 6 to 12 workflows, score against the Salesforce data foundation, sequence over 6 to 12 months. See Agentforce vs Copilot for the cross-vendor platform-selection companion.

Frequently asked questions

What is Salesforce Agentforce in one sentence?

Salesforce Agentforce is Salesforce's enterprise agent platform, launched in 2024 as the successor to Einstein Copilot, that runs autonomous AI agents inside the Salesforce platform, reasons over Data Cloud and per-cloud records, and takes action on Salesforce records (Sales Cloud, Service Cloud, etc.).

Is Agentforce the same as Einstein Copilot?

Agentforce is the evolution of Einstein Copilot. Einstein Copilot was the 2023 to 2024 brand for Salesforce's generative AI assistant. Agentforce extends that with autonomous multi-step action, agent-to-agent communication, and the broader Atlas reasoning engine. See Agentforce vs Einstein.

How is Agentforce priced?

Per conversation, where a conversation is a session of interaction (typically multi-turn). 2026 pricing lands around $2 per conversation with volume discounts. The Data Cloud subscription is separate and is usually the bigger line item for enterprises at scale.

Does Agentforce require Data Cloud?

Strongly recommended. Without Data Cloud, Agentforce reasons over per-cloud records (Sales Cloud opportunities only, Service Cloud cases only) and the cross-entity context is missing. Most production Agentforce deployments include a parallel Data Cloud investment.

What is the Atlas Reasoning Engine?

Atlas is Agentforce's reasoning engine, announced at Dreamforce 2024 and rolled out through 2025 to 2026. It is the orchestration layer that decides which tools an agent calls, how it plans multi-step actions, and how it grounds responses against the Salesforce data layer.

How does Agentforce compare to Microsoft Copilot?

Agentforce runs inside the Salesforce platform and reasons over Salesforce data. Microsoft Copilot runs inside the Microsoft tenant and reasons over Microsoft Graph plus connected sources. Most Fortune 500 enterprises run both, with Agentforce for Salesforce workflows and Copilot for Microsoft-stack workflows. See Agentforce vs Copilot.

What kinds of agents can I build with Agentforce?

Customer-facing service agents (case triage, knowledge retrieval, draft response), sales agents (opportunity research, next-best-action, pipeline cleanup), commerce agents (product recommendations, order management), and internal employee agents (HR, IT helpdesk, internal knowledge).

How does Thinklytics work on Agentforce?

We scope Agentforce engagements with workflow-first methodology, not platform-first. Map the 6 to 12 workflows the company actually wants to agent-ify, score against the Salesforce data foundation, sequence over 6 to 12 months. See Agentforce vs Copilot for the platform-selection companion.