You pay a tier uplift on everyone and a consumption pool on top
Most AI add ons meter one way or the other: a per seat subscription, or consumption. Now Assist does both. The tier upgrade lifts the underlying licence across the whole population whether or not a given user ever invokes an AI feature, and the action unit pool separately meters what actually gets consumed. Understanding that the two are independent is the entire commercial insight.
Prepared by Redress Compliance · August 10, 2026 · ServiceNow advisory. Based on 20 to 30 Now Assist pricing reviews, 2024 to 2025.
Executive summary
The tier upgrade adds 30 to 60 percent on top of the underlying licence, applied to the licensed population rather than to AI users.
Now Assist is not an add on SKU bought per adopter: the enhanced tier replaces the base tier for that workflow product, so every licensed user in it carries the uplift.
With adoption running at 30 to 40 percent in our file, the majority of the population being charged the premium never invokes the capability it pays for.
The action unit pool is a separate meter, and quoted credit packs ran 25 to 50 percent above first year actual usage. The vendor supplied consumption forecast was rarely tied to real ticket volume, so the pool was sized against an adoption curve rather than against a workload.
That is the same overcommitment pattern seen across consumption based AI generally, and here it sits alongside rather than instead of the seat premium.
The two errors compound rather than offset. A buyer paying the uplift across a population where a third use the feature, while also holding credit packs sized a third to a half above real burn, is overpaying on both meters simultaneously for the same under adoption.
Sizing one correctly does nothing for the other, because one is a per user premium and the other is a consumption prepayment.
Coverage is per workflow product, which makes the platform framing misleading. An enhanced tier on one workflow does not extend the capability to the others, so an estate running three workflow products faces three separate uplifts rather than one platform level AI decision.
And because the AI features are bundled into platform SKUs that resist clean unbundling, the moment to separate them is at purchase rather than at renewal.
What the enhanced tier covers, product by product
| Workflow product | Typical uplift on base | Action allowance included |
|---|---|---|
| Service management | 30 to 40 percent | 5,000 to 25,000 actions per user per year |
| Customer workflows | 30 to 40 percent | 5,000 to 25,000 actions per user per year |
| HR workflows | 30 to 40 percent | 5,000 to 25,000 actions per user per year |
| Creator workflows | 30 to 35 percent | 3,000 to 15,000 actions per user per year |
| Security workflows | 35 to 45 percent | Custom allocation |
Two tiers exist and they differ by more than allowance, which matters for scoping rather than only for price. The first enhanced tier adds generative capability, meaning summarisation, drafted replies, and knowledge generation, and suits workflow products where those are the primary use cases.
The higher tier adds agent and process automation, meaning components that close cases and trigger workflows rather than assisting a human who does, and carries roughly double the action allowance.
That distinction should drive the tier choice per workflow product: an estate wanting drafting assistance in one area and genuine case deflection in another has two different requirements and should not buy the higher tier across both.
The platform pricing context sits in the ServiceNow pricing guide.
Sizing each meter against what it actually measures
- Size the tier uplift against the population that will use AI, not the licensed population, and phase the tier by workflow product rather than applying it across the platform in one step.
- Size the credit pool against measured ticket volume, since the vendor forecast was rarely tied to real volume and packs ran 25 to 50 percent above first year usage.
- Treat adoption as the shared variable, because a 30 to 40 percent adoption rate under sizes nothing: it means both the uplift and the pool were bought for a population that did not materialise.
- Phase the tier upgrade by workflow product, starting where the use case is strongest, since coverage does not extend across products and there is no penalty for sequencing.
- Separate the AI components from the platform SKU at purchase, because they are bundled in a way that blocks clean unbundling later, and the renewal is too late to ask.
The ServiceNow negotiation brief
The tier arithmetic, the consumption pool sizing, the workflow scoping question, and the buyer side moves at the next renewal.
Get the white paper →Why the two meters have to be argued separately
The structural feature worth understanding is that these two charges answer different questions and therefore respond to different arguments.
The tier uplift is a per user premium on the underlying licence, so it scales with how many people are licensed for that workflow product and is entirely indifferent to whether any of them ever invoke an AI feature.
The action pool is a consumption prepayment, so it scales with how much the adopters actually do.
A buyer who negotiates hard on one and accepts the other has fixed half a problem, and the half left unfixed is usually the larger one, because the uplift applies to the whole population while the pool applies only to the active minority.
Adoption is the variable that connects them, and at 30 to 40 percent it makes both purchases wrong in the same direction at once. That produces a specific sequencing recommendation.
Phase the tier by workflow product rather than buying the platform, because coverage does not travel between products and there is no commercial penalty for adding the second workflow later once the first has proved its adoption.
Then size the pool against measured ticket volume from the workflow already running, rather than against a forecast for workflows not yet enabled.
And do the unbundling work at purchase, because the AI features are packaged into platform SKUs in a way that resists separation, which means the renewal conversation starts from a bundle nobody can decompose. The renewal sequence sits in the ServiceNow renewal playbook.
- Percentile standing for your exact deal size and industry, from real closed transactions
- Scenario simulation before the call: test alternative terms and see the financial impact of each
- A negotiation playbook, talking points, and a two page executive brief on day one
What we saw across Now Assist reviews, 2024 to 2025
Across roughly 20 to 30 Now Assist pricing reviews run between 2024 and 2025, the consumption forecast the vendor supplied was rarely tied to real ticket volume:
How far quoted action unit packs exceeded first year actual consumption, sized against an adoption curve rather than a workload.
The premium added to the underlying licence, charged across the whole licensed population for that workflow product.
Three patterns recurred: quoted credit packs running 25 to 50 percent above first year actual usage, the enhanced tier uplift adding 30 to 60 percent on top of the underlying licence, and the AI features bundled into platform SKUs in a way that blocked clean unbundling.
The buyer side move is to treat the two meters as separate negotiations, phase the tier by workflow product rather than buying platform wide, size the consumption pool against measured ticket volume, and do the unbundling work at purchase because it cannot be done cleanly afterwards.
The wider library sits in the ServiceNow practice.
Your first five moves
- Separate the two meters before negotiating either, because the tier uplift is a per user premium on the whole population and the action pool is a consumption prepayment on the active minority.
- Size the tier against the population that will genuinely use AI, not against the licensed population, since adoption ran 30 to 40 percent and the uplift applies regardless.
- Size the action pool against measured ticket volume, not the supplied forecast, which was rarely tied to real volume and produced packs 25 to 50 percent above first year usage.
- Phase the tier by workflow product, starting where the use case is strongest, because coverage does not extend between products and sequencing carries no penalty.
- Do the unbundling at purchase, not at renewal, since the AI features are packaged into platform SKUs that resist clean separation afterwards. The ServiceNow practice runs the sizing with you.
Frequently asked questions
How is Now Assist licensed?
Through two charges at once. An enhanced tier replaces the underlying tier for a workflow product and carries a 30 to 60 percent uplift applied across the licensed population, and a separate action unit pool meters actual AI consumption.
Most AI add ons use one mechanism or the other; this uses both, and they are independent.
Does the uplift apply only to AI users?
No, and that is the most expensive misunderstanding. The enhanced tier replaces the base tier for that workflow product, so every licensed user in it carries the premium whether or not they ever invoke an AI feature.
With adoption at 30 to 40 percent, most of the population paying the uplift never uses what it funds.
How accurate are the quoted credit packs?
In our file they ran 25 to 50 percent above first year actual usage, because the supplied consumption forecast was rarely tied to real ticket volume.
The pool was typically sized against an adoption curve rather than a workload, which is the same overcommitment pattern seen across consumption priced AI generally.
Does buying the tier on one workflow cover the others?
No. Coverage is per workflow product, so an enhanced tier on service management does not extend the capability to customer or HR workflows.
An estate running three workflow products faces three separate uplifts, which makes the platform level AI framing misleading when the licensing is product level.
What is the difference between the two enhanced tiers?
The first adds generative capability such as summarisation, drafted replies, and knowledge generation, suiting products where assisting a human is the use case.
The higher tier adds agent and process automation that closes cases and triggers workflows, and carries roughly double the action allowance. They answer different requirements, not different budgets.
Can the AI components be unbundled later?
Not cleanly. They are packaged into platform SKUs in a way that blocks clean separation, which means the renewal conversation starts from a bundle nobody can decompose. The unbundling has to be done at purchase, while the composition of the order is still open and the vendor wants the deal.
What is the right sequence?
Phase the tier by workflow product, starting where the use case is strongest and adoption is most likely, then size the action pool against measured ticket volume from the workflow already running rather than against a forecast for workflows not yet enabled.
There is no commercial penalty for adding the second product later.
ServiceNow's New AI: Assists, Tiers, and the Meter You Are About to Sign
The assist economy read in full: 25 to 150 assists per action, pools of 1,500 to 6,000 per fulfiller by tier, the overage cliff that is negotiable exactly once, sizing from your median curve, and choosing the tier backward from evidence.
