Full narration of the briefing. Click a section heading to jump the player to that moment.
Last time we said the rep is buying a statistic from you. Today, the thing that statistic runs on. Every tier now carries a bundled AI allowance, and every tier meters what you use beyond it. So there is a sentence I want you to carry into every conversation this cycle: the tier price is the sticker, and the consumption meter is the engine.
Buyers negotiate hard on the sticker and sign the engine without reading it. Claire is going to take you through the unit, the pool and the behaviour that catches almost everybody, and none of it is complicated. It is just new, and it was not in your last contract.
The unit is the assist. Every tier carries a bundled assist pool, and anything past the pool is billed per unit. Now the weights, because this is where forecasting breaks. Agents burn twenty five, fifty or a hundred and fifty assists per action depending on the size of the action.
A Foundation ITSM fulfiller receives roughly fifteen hundred assists a year. So heavy use burns that pool about twelve times faster than light use, on identical licensing. And cross vendor benchmarks put agentic workloads at five to ten times the consumption of a single interactive prompt. Summarising an incident is cheap.
A multi step action chain is not. Same pool, wildly different draw.
Here is the framing error that costs the most. When AI is bundled, it feels included. It is not included, it is budgeted. The pool is a budget, not an entitlement, and included up to a limit is not unlimited.
Bundling did not make the AI free; it moved the cost from a line you could see to a meter you have to model, and the true up is where the difference gets billed. That is worth sitting with, because it explains why this cycle feels calmer and costs more. Bundling lowered the anxiety and raised the bill. The line item you used to argue about has become a variable you have to forecast.
Now the single most important mechanic in this session. Usage in sub production and development instances draws on the same pool as production. That is a real behaviour change from legacy licensing, where a development instance carried no incremental cost. And the consequence is exactly what you would expect once you hear it: teams have exhausted their pools without a single production user noticing.
A test harness runs harder than any human. An agent on a schedule runs whether anyone is watching. So if your platform team is doing serious agent development, they are spending your production budget, and nothing in the environment tells them that.
Which is why a seat based forecast understates the meter by construction. The pool depletes by actions, and actions scale with automation, not with headcount. You can hold your user count perfectly flat and triple your consumption by turning on one agent. That breaks the mental model most of us built over twenty years of per user licensing, where usage and headcount moved together.
They no longer do. And notice what this does to your renewal: your seat count is the number you can negotiate confidently, and it is no longer the number that determines your bill.
So what does that cost in practice? Once Now Assist and agents were in real use, overage landed at fifteen to thirty percent of tier spend, with a median of twenty two percent. That is not a rounding error on a platform contract, it is most of a year's uplift arriving through a door nobody was watching. And two more numbers.
Unused credits do not roll over by default: they forfeit at the end of the contract year unless the paper carries explicit carry forward language. And overage rates run materially above the contracted pack rate, so the cheapest assists are the ones you committed to and the most expensive are the ones you needed.
Here is the move. Before any tier discussion, ask two questions in writing. What is the bundled assist pool, stated in absolute annual units per tenant, not per user and not as described in current documentation? Documentation gets edited; order forms do not.
And what is the overage rate, in dollars per assist? Neither number is published anywhere, so if a proposal does not contain both, it is not a price, it is an estimate with your name on it. Then ask whether non production draws on the same pool, and get the answer in writing. Next time, Daniel and I show you how to size the pool properly before you sign.
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