I am looking to get a better understanding of how the consumption based billing model works in Salesforce, specifically regarding Flex Credits.
Could anyone share a high level overview of how these credits are consumed? I am particularly interested in
- Which actions burn through these credits the fastest?
- How the multiplier system (rate cards) generally applies to different types of tasks.
- Any best practices for a team that is newly transitioning to a credit based model.
Thank you in advance!
Flex Credits Overview
Salesforce Flex Credits are a consumption-based currency used to pay for Agentforce and other AI/automation features. Each org is allocated a pool of Flex Credits that gets drawn down as features are used.
How Credits Are Consumed
- Agentforce conversations – Each autonomous agent conversation consumes credits based on the number of reasoning steps (LLM calls) taken. Complex, multi-step tasks burn more credits than simple one-turn interactions.
- Einstein AI features – Predictions, summaries, and generated content (e.g., Einstein GPT for Sales/Service) each have a per-use credit cost defined in Salesforce's rate card.
- Data Cloud activations and unifications – Some Data Cloud processes also draw from the Flex Credit pool depending on your contract.
Rate Cards (Multipliers)
Salesforce publishes a rate card that maps each feature to a credit multiplier. For example, a basic agent reply may cost 1 credit, while a complex reasoning chain may cost 5–10. You can view your org's consumption in Setup > Agentforce > Usage & Credits.
Best Practices for New Teams
- Start with a sandbox org to measure actual credit consumption before going live.
- Set usage alerts in Setup so you're notified before credits run low.
- Limit agent topic scope early on—narrower topics mean fewer reasoning steps per conversation.
- Review the Salesforce Flex Credits documentation and your order form for your specific rate card.
Hope this helps clarify the model!
