Agentforce is Salesforce's platform for building autonomous AI agents that work on top of CRM data. Introduced in 2024 and expanded through the Agentforce 360 releases, it lets a Salesforce customer define agents from topics, instructions and actions, then deploy them into service and sales workflows where they resolve cases, qualify leads and update records without a human in the loop for routine paths.
The technical core is the Atlas reasoning engine, which applies inference-time reasoning and agentic retrieval rather than a single-shot prompt against retrieved documents. Salesforce's own engineering writeups position this as the mechanism that lowers hallucination and keeps responses grounded in the customer's records. Agent Script, added with Agentforce 360, gives developers a code-based way to express agent logic — a meaningful addition, since the original declarative builder alone made complex branching awkward.
Agentforce's real advantage is proximity to the data and the permission model. Actions are built from existing Flows, Apex classes and MuleSoft APIs, agents inherit the org's sharing rules, and grounding comes from Salesforce records plus Data 360. For a company already running its service desk in Service Cloud, that removes most of the integration work that standalone agent vendors require. The corresponding limitation is just as clear: outside the Salesforce ecosystem there is very little reason to choose it.
Pricing is the hardest part to evaluate honestly. Three models run in parallel — Flex Credits at $500 per 100,000 credits (roughly $0.10 per standard action, $0.15 for voice), a $2-per-conversation model, and per-user licensing that starts around $125/user/month as an add-on and climbs to roughly $550/user/month for Agentforce 1 editions. Which is cheapest depends entirely on volume and action-per-conversation ratios, and getting that estimate wrong is expensive. There is no free tier, though existing Enterprise Edition customers can get 100,000 Flex Credits at no cost via Salesforce Foundations.
Key Benefits
- Zero-integration grounding: Agents read and write live CRM records directly, with no separate sync or vector pipeline to maintain.
- Reuses existing automation: Flows, Apex and API integrations already built by the org become agent actions rather than being rebuilt.
- Inherited governance: Sharing rules, field-level security and audit trails apply to agents the same way they apply to users.
- Code or clicks: Agent Script suits engineering teams; Agent Builder suits admins — both target the same runtime.
- Consumption option: Flex Credits let teams start without committing to per-seat licences across the org.
Use Cases
- Tier-1 case deflection — Service agents answer account, order and policy questions from live records and escalate only what falls outside their defined topics.
- Inbound lead qualification — Sales agents engage new leads, ask qualifying questions, enrich the record and book meetings with the right rep.
- Post-sale account admin — Agents process routine changes such as address updates, plan changes or returns by invoking existing Flows.
- Internal employee support — Agents grounded in internal knowledge handle HR and IT questions inside Slack or the Salesforce console.
