What is Salesforce Agentforce?
Salesforce Agentforce is an enterprise AI agent platform for organisations that want AI agents to take action across CRM, service, sales, marketing, commerce, and internal workflows. It is designed for teams already using Salesforce that need agents connected to customer records, business rules, knowledge, and approved actions rather than a standalone chatbot.
Agentforce supports the lifecycle of building, testing, deploying, managing, and observing agents. Salesforce positions it as a platform for customer-facing and employee-facing agents that can use Salesforce data and connected external systems with governance controls.
Agentforce platform capabilities
- Low-code and pro-code building: Agentforce Builder, visual workflows, prompt tools, Flows, Apex, and MuleSoft APIs help teams configure agents and actions.
- CRM and enterprise context: agents can use Salesforce records, knowledge, metadata, and connected structured or unstructured data. Salesforce describes retrieval-augmented generation and vector capabilities for grounding responses in relevant business context.
- Workflow orchestration: agents can combine deterministic workflow steps with language-model reasoning, then take governed actions across connected systems.
- Channels and use cases: deployments can support customer service, sales development, lead nurturing, employee workflows, chat, voice, apps, and portals.
- Governance and observability: Salesforce provides controls for testing, monitoring, security, access, and operational oversight.
Salesforce Agentforce pricing
Agentforce pricing is consumption-based for many use cases. Salesforce lists Flex Credits at US$500 per 100,000 credits. A standard or custom Agentforce action is listed at 20 Flex Credits, equivalent to US$0.10 per action at that rate. Salesforce also publishes conversation-based pricing of US$2 per conversation for applicable use cases.
Budgeting should include the actions each agent performs, expected request volume, relevant Salesforce editions and add-ons, Data 360 or other consumption services where used, and any implementation or integration work. Salesforce's examples are illustrative, so buyers should confirm current entitlements, regional prices, product availability, and contract terms directly with Salesforce.
Models, data, and integrations
Agentforce is an orchestration platform rather than a single-model product. Its value is in combining language-model reasoning with Salesforce business context, retrieval, workflow logic, and action execution. Salesforce also documents support for open integration approaches including Model Context Protocol (MCP) and agent-to-agent (A2A) patterns, alongside Salesforce-native tools such as Flow, Apex, MuleSoft, and Data 360.
For an evaluation, ask which models and features are available in your Salesforce edition and region, what data is available to each agent, how actions are approved and logged, and how usage is measured before rollout.
Best fit for Salesforce Agentforce
Agentforce is strongest for organisations with meaningful Salesforce adoption that want to automate sales, service, commerce, or employee processes with access to CRM data and established business workflows. It is less compelling for a small team that only needs a generic website chatbot or does not rely on Salesforce as a core system of record.
Evaluation checklist
- Start with a narrow, measurable workflow such as case triage, lead qualification, knowledge answers, appointment scheduling, or sales coaching.
- Map every agent action, system permission, escalation path, and human handoff before production deployment.
- Estimate Flex Credit usage from expected actions and volumes; monitor costs with the relevant Salesforce usage tools.
- Test response grounding, data access, guardrails, and failure cases with realistic customer and employee scenarios.