Full narration of the briefing. Click a section heading to jump the player to that moment.
Underneath the subscription run two consumption meters, and they behave nothing like the user metric you spent session four on. The platform credits that pay for extensions and integration, and the AI Units that meter Business AI. Both are prepaid balances rather than entitlements, both expire, and both are priced at the order form rather than at consumption. Which produces the sentence for this session: the seat price is fixed at signature and so is the meter rate, but only one of those is in the conversation.
Daniel and I are going to make sure both of them are, because the meters are where the surprises live.
Start with the platform. BTP sells capability as prepaid credits, not as named licences, and there are two doors. The CPEA commit gives a discount for an annual spend commitment. Pay as you go charges actual use with no commitment and no discount.
Most enterprises sit on CPEA. And the mechanic that catches people is simple: unused CPEA balance is lost at the anniversary. It does not roll. So overcommitting is not a timing problem you can recover from later, it is a direct write off, repeated every year of the term, on a number somebody forecast once before anything was built.
And the size of the miss is consistent. Across the BTP reviews we benchmarked, the average committed credit balance ran thirty to forty five percent above actual annual consumption. Buyers committed to a three year ramp based on a year one forecast, then carried that idle credit every single year. Two services in particular consumed credits two to three times faster than the customer modelled, which is Integration Suite and Analytics Cloud, so the forecast is wrong in both directions at once: too much committed overall, and the wrong mix inside it.
Right sizing beats discounting here. A twenty percent smaller commit usually saves more than five points of discount.
Then the ratchet. Mid term overage converts to a higher committed baseline at renewal, and in seven of ten renewals that is exactly what happened: the true up reset the committed baseline upward and locked the overage in permanently. Notice the asymmetry. Consume more than you committed and the baseline rises.
Consume less and the balance simply expires. The meter only moves one way, which is why the negotiation is not about the rate at all, it is about the commit level, the rollover language, and whether a true up is allowed to reset the floor. Ask for that last one explicitly, because silence on it is a yes to them.
The AI meter works the same way with newer numbers. An AI Unit is a prepaid credit drawn per action across the tenant whenever Business AI runs beyond the bundled base tier, and the draw scales with task complexity: agents draw far more than interactive prompts. Roughly two hundred AI actions are bundled per Advanced FUE and they pool across the estate, which looks like months of headroom at chat volumes and can be spent in weeks by a few scheduled agents. Overage ran about eight to eighteen cents per action in the quotes we reviewed, and in our renewals that overage rate was rarely negotiated at all.
Now the good news, because one of these meters moves strongly in your favour. The platform credit pool opened at three to five percent of annual contract value and closed at eight to twelve percent once the buyer brought a list of named extensions, integrations and services with owners attached. It roughly triples, and it is won with a document rather than with a negotiating position. On a large agreement that swing is worth more in absolute terms than several points of rate discount, and it is exactly the line buyers tend to accept as a throw in.
Bring the use case list. It is the cheapest leverage in the whole negotiation.
Here is the move, and it is three sentences on the order form. Forecast bottom up from the agents and integrations actually on your roadmap rather than from today's chat volume, because the meter never counted users in the first place. Cap the overage rate in dollars per action and hold it for the term. And write that a true up may not reset the committed baseline at renewal.
Then bring the documented use case list to triple the credit pool. Next time, Daniel and Claire take the year that quietly decides whether this deal saved you anything at all: year four.
Redress Compliance works on contingency: our fee is 25 percent of what we save you. Nothing saved, nothing paid. Independent, buyer side only, never vendor funded. Want Redress to contact you? Reach out and we respond the same day.
Talk to a ServiceNow negotiator