Now Assist is ServiceNow's generative-AI layer, and it is billed per assist — every metered AI action draws down a consumption pool. The Admodum read on how per-assist pricing works, why the cost scales with adoption rather than seats, and how to scope the pool and cap the unit rate before the feature is switched on.
Now Assist is ServiceNow's generative-AI layer, and it is priced per assist rather than per seat. Admodum is an independent, buyer-side software licensing advisory firm; this page explains the per-assist model and the deal mechanics around it because, of every line on a modern ServiceNow contract, the AI consumption line is the one most able to grow without warning.
An assist is a single discrete generative-AI action — summarising an incident, drafting a resolution note, generating a knowledge article or producing code inside a flow. Each assist draws down a pre-purchased consumption pool, a quantum of assists bought up front in the same way a buyer might pre-purchase any metered resource. The model is fundamentally different from the seat-based licensing that governs the rest of the platform, where a known headcount produces a known cost. Here the cost is a function of behaviour, and behaviour is far harder to forecast than headcount. The seat-based mechanics this departs from are set out at the ServiceNow licensing model pillar, and the wider renewal context at ServiceNow renewal and negotiation.
Because an assist is billed on use, the cost of Now Assist rises directly with adoption. A successful internal rollout — the very outcome the platform team is rewarded for — is precisely what drives the consumption line upward, which inverts the usual relationship between a feature working well and its cost staying flat.
This is the structural trap in consumption pricing. Under a seat model, encouraging more people to use a tool they are already licensed for is free; under a per-assist model, every additional habit formed is a recurring charge. As agents learn to lean on summarisation, as virtual-agent deflection scales, and as automated flows begin invoking generative actions of their own, the assist count climbs on a curve the original forecast rarely anticipated. The danger is not that Now Assist fails but that it succeeds, and the bill arrives sized to that success.
An uncapped consumption pool can be exhausted well before the term ends, at which point the contract's overage mechanism takes over. If the per-assist overage rate was not negotiated down in advance, the excess is billed at list, and the buyer discovers the true unit cost of the feature only after it has been incurred.
Two failure modes recur. The first is the silent overage: consumption crosses the pool ceiling mid-term and continues billing at an unfavourable rate nobody is watching, surfacing only at the next invoice or true-up. The second is the renewal shock, where a year of unmonitored adoption is reconciled at once and the buyer is asked to fund a much larger pool on the vendor's terms, framed against a deadline. Both are avoidable, but only with the contractual scaffolding put in place before the feature goes live — the same discipline applied to the seat-based usage analytics and the annual true-up, and to the renewal uplift caps that bound the rest of the agreement.
Controlling Now Assist cost rests on three levers, all negotiated before the feature is enabled: a right-sized consumption pool, a capped per-assist overage rate, and clear contractual definitions of which actions consume the pool. Pulled together they convert an open-ended consumption liability into a bounded, monitorable one.
The pool size should follow a realistic adoption forecast rather than the vendor's optimistic projection or the buyer's nervous underestimate; too small invites overage, too large is shelfware bought up front. The overage rate must be capped, so that if consumption does exceed the pool the excess is billed at a known, pre-agreed price rather than at list discretion. And the definition of a billable assist must be explicit, so the buyer is not metered for actions triggered automatically by background processes or for retries the user never intended. With those three in place, the remaining task is simply to monitor actual consumption against the pool through the term, which removes the renewal shock before it can form. The clause detail behind the protections sits at price protection, co-term and swap clauses, and the optimisation that frees budget for the AI line at entitlement optimisation and shelfware recovery.
Now Assist is best scoped as part of a wider renewal, not bought mid-term as an isolated add-on. Folded into the renewal, the AI consumption line can be traded against the rest of the contract value; bought alone, it carries almost no negotiating tension and tends to be priced at or near list.
The reason is leverage. During a renewal the buyer has commitments the vendor wants — term length, module retention, expansion — and the Now Assist pool, unit rate and protection clauses can be set against those. Outside a renewal, the buyer is a willing purchaser with a deadline of their own making and little to trade, which is the weakest position from which to size a consumption pool. Where the platform team is impatient to switch the feature on, the advisory task is to hold that enthusiasm until the commercial moment is right, then scope the AI line properly as one element of the renewal rather than as an afterthought to it. The full method sits at the ServiceNow knowledge hub and the ServiceNow blog cluster; the engagement opens at the ServiceNow practice or directly at contact.
Now Assist is priced per assist rather than per seat. An assist is a single discrete generative-AI action, such as summarising an incident or drafting a resolution note, and each one draws down a pre-purchased consumption pool. Because billing tracks usage rather than headcount, the cost of Now Assist rises with adoption, and a successful internal rollout is precisely what drives the line item up.
An assist is the billing unit for Now Assist: one metered generative-AI action invoked by a user or a workflow. Summarising a case, generating a knowledge article, drafting a chat reply or producing code in a flow can each count as an assist. The definition matters because the contract should state exactly which actions consume the pool, so the buyer is not billed for assists triggered automatically by background processes.
Because consumption scales with adoption, not with a fixed seat count, an uncapped Now Assist pool can be exhausted faster than the buyer forecast, triggering an overage charge or a mid-term true-up at a unit rate that was never negotiated. The feature is designed to be used heavily, so the more value the platform team delivers, the larger the consumption bill, unless the pool size and the overage rate were scoped in advance.
Control rests on three levers negotiated before the feature is enabled: a right-sized consumption pool based on a realistic adoption forecast, a capped per-assist overage rate so any excess is billed at a known price rather than list, and clear contractual definitions of which actions consume the pool. Monitoring actual assist consumption against the pool through the term then prevents a surprise at renewal.
Scoping Now Assist as part of a wider renewal gives the buyer more leverage than buying it mid-term as an add-on, because the AI line can be traded against the renewal value rather than priced in isolation at list. A mid-term purchase carries little negotiating tension; folding the consumption pool, the unit rate and the protection clauses into the renewal lets the buyer set the terms while other commitments are on the table.
A senior Admodum ServiceNow advisor will forecast your assist consumption, right-size the pool, cap the overage rate and fold the AI line into the renewal. The Now Assist scope white paper sets out the method; renewal moments route to the Renewal Programme and the monthly read sits in the newsletter.