Here is the year-one story we see most often. A ServiceNow buyer is sold on agentic AI, runs the numbers off the vendor's projected adoption, and commits to a generous annual assist pool to capture the best unit price. Twelve months later, adoption has reached perhaps a third of the forecast, a large slice of the pool has gone unused and expired, and the renewal team opens negotiations using last year's commitment as the baseline. The buyer paid for capacity twice over: once for the unused units, and again because the inflated number now sets the renewal floor. None of it was necessary.
Why do buyers overbuy assists in the first year?
Because there is no usage history to size against, and a vacuum gets filled with optimism. The vendor models a steep adoption curve, the buyer wants the volume discount that comes with a bigger commitment, and the two pressures push the year-one number up. The problem is that AI adoption inside a service desk ramps on human behavior, not on license availability. Agents have to trust the tool, workflows have to be configured, and end users have to change habits. That takes quarters, not weeks.
For the full picture of how ServiceNow is pricing this, including the Foundation, Advanced and Prime consolidation with AI bundled from April 2026, see our 2026 guide to agentic AI ITSM pricing and negotiation.
Book an AI cost control review →How are ServiceNow assists actually consumed?
Assists are metered units of AI work, and different actions consume very different amounts. As a rough guide published around the 2026 model, a small action runs about 25 assists and a large agentic action about 150, with overage handled through top-ups. That spread is exactly why year-one estimates go wrong: a handful of heavy agentic workflows can consume more than a large volume of light summarization. You cannot size a pool credibly until you have seen which mix your team actually runs. The mechanics are detailed in how Now Assist assists and tokens work.
What does a right-sized first year look like?
| Approach | Year-one risk | Renewal effect |
|---|---|---|
| Large upfront pool for best unit price | Unused units expire; you pay for idle capacity | Inflated baseline anchors the renewal high |
| Vendor-forecast pool | Forecast assumes adoption you have not earned yet | Shortfall or surplus, both priced against you |
| Measured ramp with capped expansion | Minimal; you buy what you use plus a buffer | Real usage sets a defensible renewal number |
What should you negotiate to stay protected?
- Start from a measured baseline. Run a short pilot or first-quarter measurement before committing the annual number. Size from observed consumption, not the forecast.
- Lock the expansion price. Fix the unit rate for additional assists and the next tier up, so growth is cheap and predictable when usage arrives.
- Cap and convert overage. Negotiate overage top-ups at the same unit rate, not a penalty rate, and convert sustained overage into the committed tier at the locked price.
- Protect the renewal baseline. Agree that the renewal is sized on actual consumption, not on the prior commitment, so an over-buy cannot anchor you high.
- Keep the term short on the AI line. A one-year AI commitment inside a longer platform deal lets you re-price once you have real data.
If you want to cap consumption at the contract level rather than just manage it operationally, the clause-level approach is in how to cap ServiceNow assist consumption before you sign.
Our gated ServiceNow Now Assist Cost Control Guide includes the assist-sizing worksheet and the overage and expansion clauses that keep year-one consumption honest.
How much does getting this right save?
It varies by deal size, but the saving is rarely a single discount. Across more than 500 ITSM engagements we average a 30% reduction, and on AI lines the bulk of it comes from not committing to capacity you will not use and from locking the expansion price so the vendor cannot re-rate growth at renewal. A buyer who buys what they consume and grows into a fixed rate keeps the savings; a buyer who buys the forecast funds the vendor's optimism.
What to do next
Before you sign, get a measured read on likely first-year consumption and the expansion and overage terms that protect you. We run this end to end through our AI cost control service on fixed fee or gainshare, and the platform-level commercial context sits on the ServiceNow platform page.
Frequently asked questions
- Why do buyers overbuy assists in year one?
- The vendor sizes the commitment from an optimistic adoption curve with no real usage to check it against. Adoption ramps slower, the unused pool expires, and the inflated number anchors the renewal. Size small and stage the rest.
- How are ServiceNow assists consumed?
- As metered units, roughly 25 for a small action and 150 for a large agentic action, with overage top-ups. Because heavy workflows burn many assists, year-one estimates are easy to get wrong without measured usage.
- Should you buy a large pool for a better unit price?
- Usually not in year one. A deeper discount on a pool you cannot consume loses to a smaller commitment with a locked expansion rate. Grow into it once usage is real.
Right-size your AI commitment.
We size your first-year assist pool from real volume and lock the expansion and overage terms that protect the renewal. Fixed fee or gainshare.
Book an AI cost control review →