The Pricing Model Change That Actually Matters
Salesforce quietly changed the commercial terms for Agentforce Help Agent when it launched June 25. The new model: $2 per issue, charged only when the agent resolves a case autonomously. Escalations and unresolved cases cost you nothing.
That is a structurally different deal than what most Salesforce customers are used to. Per-conversation billing — the prior model — charged regardless of outcome. You absorbed the cost of every failed interaction. Pay-per-resolution flips that. Salesforce now has a financial stake in whether the thing actually works.
The internal data they're citing is credible enough to take seriously: 4.3 million inquiries handled on help.salesforce.com, 70% resolved autonomously. That's not a pilot. It's a production signal.
What to Model Before July GA
If your organization is on Flex Credits or a per-conversation plan, you have roughly four weeks to run a cost comparison before the Help Agent goes generally available. The math is straightforward but the inputs matter:
- Current cost per conversation (including unresolved and escalated cases)
- Your actual autonomous resolution rate — don't use Salesforce's 70% benchmark; use your own support ticket data to estimate what a well-configured agent would realistically close
- Volume by case type — resolution rates vary sharply by issue complexity; segment before you aggregate
At 50% autonomous resolution, the effective cost per case interaction is $1. At 70%, you're paying for a fraction of what you previously spent on volume. At 30%, the model still protects you from paying for failures — which is the floor that makes this worth evaluating even under pessimistic assumptions.
The Procurement and Governance Play
Outcome-based pricing materially lowers the internal hurdle for budget approval. When the vendor absorbs the cost of failure, the risk profile of the purchase changes — and that changes how you should position this with finance and legal.
The pre-wired integration with Salesforce Knowledge also matters operationally. Setup friction is a real reason AI deployments stall in non-tech organizations. Removing the knowledge-base integration step shortens the path from contract to production.
The actionable move now: pull 90 days of support ticket volume, segment by case type, apply a conservative resolution-rate assumption, and run the comparison against your current per-conversation spend. Have that model ready before the July GA conversation with your Salesforce account team. Vendors negotiate differently when the customer walks in with the math already done.