Short version: consumption-priced AI, whether assists, tokens or resolutions, includes a bundle, and everything past the bundle is overage at a higher unit rate or a top-up you pre-buy. Vendors quote the bundle and stay quiet on the overage rate, because that is where the bill grows after you have signed and lost your leverage. The fix is to treat the overage rate as a first-class negotiated term, not a footnote. For the wider pricing picture these charges sit inside, start with our 2026 guide to agentic AI ITSM pricing and negotiation.
Book an AI cost control review →What are AI overage and top-up charges?
Overage is the charge for consumption beyond what your subscription includes, billed per unit at a rate that is usually higher than the effective rate inside your committed bundle. Top-ups are pre-purchased blocks of extra consumption you buy in advance, often at a slightly better rate than reactive overage but with the risk you pay for capacity you never use. As of 2026, ServiceNow meters Now Assist in assists, roughly 25 for a small action and about 150 for a large agentic action, with overage top-ups when you exceed the commitment. The units differ by vendor, but the shape is the same: a committed bundle, a higher reactive rate, and a pre-buy option in between.
Why it matters to you: AI usage tends to rise as you switch on more workflows, so a contract that looks fixed at signature drifts into variable overage by mid-term. If the overage rate is not capped, the vendor sets the marginal price of your own success. The consumption mechanics behind this are worth reading in full in per-resolution vs per-seat AI ITSM pricing.
How do overage and top-ups compare?
They are two ways to pay for the same thing past your bundle, and the right mix depends on how predictable your usage is. Here is the trade-off.
| Mechanism | How it bills | Risk to you | Best when |
|---|---|---|---|
| Committed bundle | Fixed, included in subscription | Paying for unused capacity if over-sized | Usage is known and stable |
| Top-up (pre-buy) | Pre-purchased block, often mid-rate | Stranded capacity if demand is lower than bought | Usage is rising but roughly forecastable |
| Reactive overage | Per-unit, highest rate, billed after use | Uncapped marginal cost, surprise invoices | Short, occasional spikes only |
Your move: aim to live inside a right-sized bundle for steady-state demand, use top-ups for planned growth, and treat reactive overage as the exception you have capped, not the default you accept.
How do you cap overage before you sign?
Run these five moves in order. Each one closes a gap the vendor would otherwise leave open.
- Fix the overage rate for the term. A not-to-exceed per-unit rate, held flat for the contract, so success does not get repriced against you.
- Replace auto-billing with a true-up. Require notification at a usage threshold and a periodic true-up, so you are never surprised by an invoice and can act before the meter runs.
- Win a conversion right. If overage is sustained, you can convert it into a re-priced commitment at the lower bundle rate rather than paying the overage premium indefinitely.
- Cap total AI growth. Put a ceiling on the percentage your annual AI spend can rise, so the variable line cannot outrun your budget.
- Get audit-grade visibility. A usage dashboard and the right to reconcile, in writing, so you are never billed off a number you cannot verify.
For the platform-specific version of capping consumption, our piece on how to cap ServiceNow assist consumption before you sign walks the assist meter line by line, and pairs well with the per-employee model covered in the Espressive Barista pricing guide when you are comparing pricing shapes.
Should you pre-buy top-ups?
Only after a metered pilot tells you what you actually consume. Pre-buying before you have usage data is guessing with your budget: too high and you strand capacity, too low and you fall into reactive overage at the premium rate. Run a short, metered pilot, measure real assists or resolutions per workflow, then size the commitment to observed demand with headroom you control rather than headroom the vendor sells you. Your move: never let a pilot's enthusiasm set your commitment; let the pilot's meter set it.
Our gated Now Assist cost control guide includes the overage clause language and the assist-forecasting worksheet behind this sequence.
What to do next
Before your next AI renewal or net-new signature, pull your consumption data, model year-two demand not year-one, and bring a fixed overage rate, a true-up process, a conversion right and a growth cap to the table. If you have no usage history yet, run a metered pilot first and size the commitment off the meter. Done in this order, overage stops being a surprise and becomes a number you set.
When you want this handled for you, it runs through our AI cost control service on fixed fee or gainshare. We have negotiated more than $420M in ITSM contract value across 500+ engagements since 2019, and consumption-based AI terms are now a core part of that work.
Frequently asked questions
- What are AI overage and top-up charges?
- Overage is per-unit billing for consumption beyond your bundle, usually at a higher rate; top-ups are pre-purchased blocks of extra capacity. Both turn a fixed AI cost into a variable one.
- How do you cap overage in a contract?
- Fix the overage rate for the term, require a true-up instead of auto-billing, win a right to convert sustained overage to the bundle rate, cap annual AI growth, and secure audit-grade usage visibility.
- Should you pre-buy top-ups?
- Only after a metered pilot shows real consumption. Size the commitment to observed demand with headroom you control, not to the vendor's forecast.
Book an AI cost control review.
We cap the overage rate, build the true-up, win the conversion right, and stop your AI line running away. Fixed fee or gainshare.
Book an AI cost control review →