The Double-Billing Problem
Salesforce's emerging pricing model for Agentforce and MCP-based integrations introduces Flex Credits as the metered cost for third-party AI agents calling into the platform. On paper, that sounds like a reasonable consumption model. In practice, it has a structural flaw: those same API calls already exist, and organizations can execute them today without triggering Flex Credit charges.
A Salesforce Technology Engagement Manager said it plainly in a Salesforce Ben analysis published this week: teams can call normal APIs and get the same result without paying Flex Credits. That's not a fringe opinion from a dissatisfied customer — it's an architectural observation from someone whose job is to implement these systems at scale.
What This Creates in Practice
When a governed, Salesforce-preferred path costs more than an unstructured workaround that produces the same outcome, rational architects will take the workaround. That's not bad behavior — it's basic incentive math.
The risk for operators isn't just the cost delta. It's that workarounds accumulate. Teams that route around MCP and Flex Credits to avoid a consumption surcharge end up with integrations that sit outside Salesforce's governance layer — harder to audit, harder to hand off, and harder to untangle when the contract comes up for renewal. The vendor's pricing model inadvertently subsidizes technical debt.
Salesforce has a long history of adding pricing tiers that create parallel architecture patterns inside the same org. AI-era consumption models amplify that dynamic because the surface area is larger and the stakes — data lineage, agent behavior, compliance — are higher.
What to Do Before You Sign Anything
If your organization is currently evaluating Agentforce or any MCP-based integration, run two cost models before committing to a consumption-based contract:
- Scenario A: Full MCP/Flex Credit path as Salesforce recommends it.
- Scenario B: Equivalent outcome using existing API calls under your current license terms.
Model both against your anticipated call volume at 6, 12, and 24 months. The gap may be negligible — or it may be significant enough to change your architecture decision entirely.
Also document which path your architects actually prefer and why. If the answer is "we'd use legacy APIs to avoid the surcharge," that preference should surface in your vendor negotiation, not emerge silently post-implementation.
Pricing that discourages the governed path is a vendor design problem. It shouldn't become your technical debt.