Salesforce's new Agentforce editions (Core at $195/user/month, Advanced at $395, Max at $550) each come with an org-level annual pool of Flex Credits: 500,000, 1 million, and 2.75 million respectively. HubSpot moved its Breeze Customer Agent to outcome-based pricing in April 2026, charging 50 HubSpot Credits per resolved text conversation at a list rate of $0.01 per credit. Two of the biggest platforms in B2B ops rewired their billing in the same year, and neither change is a simple price increase. Both add a second meter on top of the seat you already pay for.
That's the part worth paying attention to. Not the number on the invoice (yet), but the structure underneath it.
The Double-Meter Problem
For years, SaaS pricing was seats plus maybe a storage tier. Predictable enough that RevOps could model renewal costs in a spreadsheet without breaking a sweat. The new structure layers consumption on top: you keep the seat license, you keep the platform fee, and now you also pay per action, per resolution, or per credit pool overage.
Salesforce's own pricing observatory shows Service Cloud Enterprise list prices moved from $165 to $175/user/month (effective August 2025), a roughly 6% bump. Then came the Agentforce editions announced in September 2026, bundling Flex Credit pools at the org level at a list rate of $500 per 100,000 credits. Conversation pricing exists as an alternative at $2 per conversation, depending on the agreement. The renewal conversation now involves at least three line items: seat, access add-on, and consumption.
HubSpot's move is slightly different. They shifted from $1.00 per conversation to $0.50 per resolved conversation, which can actually lower costs for teams whose resolution rates are high. Monthly included credits range from 500 (Starter) to 5,000 (Enterprise). But those credits are shared across features, not reserved for the Customer Agent alone. If your other Breeze features consume credits first, your agent budget shrinks without anyone changing a setting.
Forrester has flagged this directly: consumption-based AI pricing shifts forecasting risk back onto the customer.
Why Org-Level Credit Pools Create Internal Budget Fights
Salesforce's Flex Credit pools are allocated at the org level, not per team. Marketing, sales, and service draw from the same bucket. If service spins up a high-volume customer-facing agent, it can burn through the pool before marketing's campaign automation agent runs its first batch.
This is a governance problem before it's a cost problem. Without internal allocation rules, the team that moves fastest gets the credits. The practical question for RevOps: who owns the credit budget? If nobody does, the org-level pool turns into a tragedy of the commons. If someone does, you've just created a new internal approval workflow for every agentic use case.
The Spiral Nobody Plans For
Agentic workflows, by design, generate more events than human workflows. An agent reading records, checking statuses, and triggering actions can produce 30,000+ API calls per day on a single integration. Every one of those calls is a billable event under the new meters. The automation you built to reduce headcount costs now increases your variable platform costs, potentially faster than it increases pipeline or retention.
That's the spiral: automation drives higher consumption, consumption drives higher bills, and the ROI calculation for the agent itself degrades. If the cost of running the agent grows proportionally with its activity (not with the value it creates), you hit a point where the agent's marginal cost exceeds its marginal contribution.
IDC has pushed back on catastrophist framing (they called the "SaaSpocalypse" narrative a flawed premise), and they're right that this isn't a collapse. But the risk is specific: unmonitored consumption meters can turn a fixed-cost tool into a variable-cost liability without anyone noticing until the renewal lands.
What to Actually Do Before Your Next Renewal
The meter itself isn't the problem. Agent traffic is real compute, and someone pays for it. The problem is the additive layering: seat plus access plus consumption, with no mechanism to substitute one for another. Three things to put in place now:
- Line-item forensics on every renewal. Compare old order forms against new ones, line by line. Look for renamed SKUs, new "AI access" add-ons, and changes to what counts as metered vs. unmetered. Salesforce's own help documentation notes that what is unmetered vs. metered depends on the specific add-on/edition and use case (employee-facing vs. customer-facing agents). If you don't know which bucket your agents fall into, the vendor does.
- Internal credit allocation model. If your vendor uses org-level pools, build a simple allocation framework before the credits go live. Assign budgets per team per quarter. Set a utilization threshold (say 80%) that triggers a review, not an overage purchase.
- A cost-per-outcome metric that spans the full bill. Don't track agent cost per resolution in isolation. Track the all-in cost: seat license + agent access SKU + consumption charges, divided by successfully handled issues. That's the number you negotiate against.
Vendors that publish the rate, cap the overage, and let an agent replace a seat (rather than stack on top of one) will hold their position. The ones that leave consumption vague and uncapped are betting that switching costs absorb the frustration.
The real question for ops isn't whether these prices are fair. It's whether your automation is generating billable events faster than it's generating qualified pipeline. If you can't answer that with a number, the renewal conversation is going to be uncomfortable, and the vendor will have better data than you do.
