ServiceNow's published overage rate of $0.20 per assist is an anchor, not a price: negotiated top-ups land between $0.02 and $0.10, a 2x to 10x spread on the same unit
ServiceNow publishes no universal per-assist price, so the $0.20 rate card exists to frame your quote rather than govern it. The AI Starter Pack gives you the one public reference point that converts into a defensible target, and the 500,000-assist adoption bonus proves ServiceNow will move 77 percent off sticker when it wants adoption. Price the unit before you argue about volume, because a bad rate applied to a good forecast still costs you six figures over a three-year term.
Prepared by Redress Compliance · August 28, 2026 · ServiceNow advisory. ELA and Now Assist renewal engagements, 2024 to 2026.
Executive summary
The $0.20 published rate is a negotiating anchor with no floor behind it, and advisory benchmarks put settled top-up rates at $0.02 to $0.10 per assist.
That spread means two buyers with identical consumption profiles can pay ten times apart for the same units, and the difference is entirely a function of when and how hard the rate was negotiated.
The AI Starter Pack is the only public price-to-volume reference ServiceNow has put in writing: 25 Pro Plus users at 6,000 assists each, 150,000 assists, plus a free 500,000-assist bonus if consumed inside six months.
A buyer who triggers that bonus receives 650,000 assists for the price of 150,000, an effective 77 percent discount that ServiceNow itself designed and published.
Agentic tiering at 25, 50 and 150 assists per execution turns the unit rate into a per-workflow price of $5.00, $10.00 or $30.00 at card rate.
A single large agentic workflow run 2,000 times a month costs $720,000 a year at $0.20 and $72,000 at $0.02, which is why the unit rate deserves more negotiating time than the volume forecast.
The April 9, 2026 collapse from five tiers to three, with legacy SKU end of sale on July 1, 2026, removed the "stay on Pro" fallback and manufactured the top-up market.
Every tier now carries a bundled pool ServiceNow does not publish, so buyers cannot compute an implied rate from the tier and must extract the top-up rate as a separate contractual number.
What the assist unit actually is, and where the price hides
An assist is not a query, a conversation, or a user. It is ServiceNow's unit of metered work, and its dollar cost is set by four variables that get negotiated separately and are rarely reconciled by the buyer before signature.
First, the published card rate: Raymond James documents overage at $0.20 per assist. Second, the negotiated top-up rate, which our benchmarking puts at $0.02 to $0.10, a 2x to 10x spread against the card. Third, and this is where most cost models break, consumption weighting per action type.
An incident summarization burns 1 assist. App creation burns 20. Agentic executions burn 25, 50, or 150 depending on tool count (0 to 4 tools, 5 to 8, 9 to 20 per the June 2026 schedule).
Fourth, pool mechanics: entitlements pool at the account level across production and sub-production, track on a rolling 365-day basis, and do not roll over. Sub-prod testing burns production budget. Year one surplus dies at reset.
If you settle the rate but ignore the weighting table, you have negotiated a discount on the cheapest action in the catalog and a blank check on the most expensive one.
| Action type | Assists consumed | Cost at $0.20 card | Cost at $0.05 negotiated |
|---|---|---|---|
| Incident summarization | 1 | $0.20 | $0.05 |
| App creation | 20 | $4.00 | $1.00 |
| Agentic execution, 0 to 4 tools | 25 | $5.00 | $1.25 |
| Agentic execution, 5 to 8 tools | 50 | $10.00 | $2.50 |
| Agentic execution, 9 to 20 tools | 150 | $30.00 | $7.50 |
The table shows the arithmetic. It cannot show the margin story sitting underneath it. Raymond James confirms buyers pay the same per-action rate regardless of which underlying model or LLM infrastructure ServiceNow uses. Inference costs have fallen sharply and will keep falling across your term.
Every cent of that decline accrues to ServiceNow unless your contract says otherwise.
There is a second hidden multiplier the price sheet never mentions. A fractured CMDB or a low Article Quality Index does not produce a failed assist, it produces retries, loops, and escalations, each of which consumes units. You pay your negotiated rate multiple times for one answer.
That is why the rate conversation and the data readiness conversation belong in the same meeting, and why how the pool and overage actually bill matters more to your three-year TCO than the headline uplift on your tier.
Converting the AI Starter Pack into a defensible target rate
You do not need ServiceNow to publish a per-assist price. They already published a ratio, and a ratio plus a price equals a rate. The ITSM AI Starter Pack bundles 25 Pro Plus users at 6,000 assists each, 150,000 assists total.
Take whatever your rep quotes for that pack, divide by 150,000, and you have ServiceNow's own implied unit economics on the record. That number will land nowhere near $0.20. Then make the arithmetic worse for them.
Consume the 150,000 inside six months and ServiceNow grants 500,000 additional assists at no charge, valid for a year. The correct denominator is therefore 650,000, not 150,000.
A buyer who triggers the bonus receives 650,000 assists for the price of 150,000, an effective 77 percent discount on the sticker unit. At the card rate, that free block is nominally worth $100,000.
ServiceNow gives it away voluntarily to drive adoption, which tells you precisely what they think a marginal assist costs them.
Two further facts sharpen the counter. The bonus is stackable: buy Starter Packs across multiple product lines, pool the assists, burn the combined total in six months, and you earn a 500,000 bonus per pack purchased. And the ratio itself is negotiable by product line.
The CSM Starter Pack carries 250,000 assists scoped to the same 25 fulfillers, 10,000 per user against ITSM's 6,000. ServiceNow has already conceded a 67 percent higher per-seat allocation for identical seat counts in a different catalog entry. Use it.
Your opening counter to $0.20 should be your Starter Pack quote divided by 650,000, expressed as a fixed top-up rate held flat for the full term with no annual uplift.
In our experience across ServiceNow renewals, buyers who anchor this way settle in the $0.03 to $0.06 band; buyers who accept the card rate as a starting point negotiate down to $0.12 and call it a win. Same vendor, same unit, twice the cost.
The anchor decides the outcome, and at higher spend levels the ratio itself becomes negotiable alongside the discount curve you should expect at your ACV band.
ServiceNow Price Benchmarks by Spend Tier: What Good Looks Like at 500K, 2M and 5M
The buyer side playbook for ServiceNow Price Benchmarks by Spend Tier: What a Good Number Looks Like at 500K, 2M, and 5M ACV, free behind a work email.
Get the white paper →Why the rate card exists to be broken
Start with what ServiceNow chose not to publish. There is no universal per-assist dollar rate in any public document, no tier allocation values disclosed for the three AI-native tiers that replaced the five legacy ones in April 2026, and no floor price anywhere in the paper.
What does exist is a rate card showing $0.20 per assist, surfaced through analyst channels rather than through a price list you can hold ServiceNow to. That asymmetry is the entire design.
A number with no published floor beneath it is not a price, it is a ceiling built to make everything below it feel like a concession.
When your account executive shows you $0.12 and calls it a 40 percent discount, the discount is measured against a figure ServiceNow invented for exactly that conversation.
The strongest evidence that $0.20 has no relationship to cost is ServiceNow's own giveaway. A buyer who consumes the 150,000 assists in the AI Starter Pack within six months receives 500,000 more at no charge, and the bonus stacks across product lines.
That is 650,000 units delivered for the price of 150,000, a 77 percent effective discount, and at card rate the free portion is nominally worth $100,000. No vendor gives away $100,000 of anything that costs it $100,000 to deliver.
The bonus is a confession that the marginal cost of an assist sits somewhere in the fractions of a cent, and that ServiceNow will trade the entire margin away when adoption is the objective. Your job in the negotiation is to make adoption the objective.
The model-agnostic clause compounds this. Buyers pay the same per-action rate regardless of which underlying model executes the task, which means every improvement in inference economics between now and your renewal accrues entirely to ServiceNow.
Inference costs have fallen sharply across every major model family since 2023 and continue to fall. A fixed per-unit rate against a declining cost base is not a pricing model, it is a widening margin.
If you sign $0.10 for three years, you are underwriting ServiceNow's cost curve, not your own consumption. This is the argument to put on the table directly, and it is the argument that moves rates when volume arguments do not.
The April 2026 tier collapse changed the leverage map more than most buyers have absorbed. When AI was an opt-in Pro Plus add-on, you could negotiate whether to buy it at all.
Now that an assist pool ships bundled inside every tier, the buy decision is gone and the platform uplift is priced into a per-seat number you are already committed to. The top-up rate is now the only genuinely variable price left in the AI portion of your contract.
Everything else was decided when you accepted the tier.
That makes the unit rate the last place where negotiation actually changes your three-year cost, and it explains why ServiceNow works so hard to defer that conversation to a later true-up, as our analysis of how Now Assist gets bundled into the tier and billed past the pool sets out.
Here is where the leverage sits, and it is not where most buyers look for it. ServiceNow's problem in 2026 and 2027 is not overage revenue, which is immaterial against total ACV. Its problem is proving to investors that customers are actually consuming the AI it has embedded across the platform.
Consumption is the metric that validates the story. A customer who signs a high rate, gets scared of the meter, and throttles usage is worth less to ServiceNow than a customer who signs a low rate and burns three million assists a year.
That is the trade you should be making explicit: aggressive adoption commitments in exchange for a rate that makes aggressive adoption rational. Adoption uncertainty is your asset. Do not give it away for free by pretending you are confident in your forecast.
The posture that follows is straightforward. Treat $0.20 as a marketing artifact and refuse to negotiate against it. Anchor instead on the Starter Pack economics, where ServiceNow has already demonstrated what it will accept when it wants the usage.
Price the unit before you argue about volume, because a rate you cannot move later is the one line item that compounds across every year of the term.
What ServiceNow does when you push on the unit rate
The counter-moves are consistent enough to plan for. First, ServiceNow offers a larger bundled pool instead of a lower rate, because pool volume costs it nothing while a documented low rate sets precedent across the account base. Take the pool and still insist the rate appears in writing.
Second, you get a one-year promotional rate that reverts to card at renewal, which is a discount on year one and a repricing event on years two and three. Third, and most common, the rep declines to price top-ups at all on the grounds that you will not exhaust the pool.
That is not a courtesy, it is a plan to price the unit when you have no alternative and no time. Fourth, the July 1, 2026 legacy end of sale gets used as a deadline. It is a packaging date, not a commercial one, and it does not expire your ability to hold the paper open.
| Vendor move | What it protects | Your counter | Target outcome |
|---|---|---|---|
| Bigger pool, no rate | Precedent on unit price | Accept pool, still require rate in the order form | Both, not either |
| Promo rate, year one only | Renewal repricing to card | Rate protection across full term plus first renewal | Flat rate, CPI cap at renewal |
| "We'll price it if you exceed" | Pricing you at zero leverage | Refuse to sign without the number | $0.03 to $0.05 committed |
| July 1, 2026 end of sale urgency | Deal timing control | Treat as packaging date, not deadline | No rate concession for speed |
| No true-down on unused pool | Full pool revenue regardless of use | Credit or carry language on unused units | Unused-pool credit at 100% |
The table shows four different tactics but only one underlying objective: keep the unit price out of the signed document. Every move here converts a rate negotiation into something else, a volume negotiation, a timing negotiation, a future negotiation.
The moment the number lands in the order form, ServiceNow has lost the ability to reprice you at the exact point where you have no alternative supplier and a production workload running on the meter.
A strong outcome is specific: $0.03 to $0.05 per assist committed in the order form, rate held flat for the full term with renewal uplift capped at CPI, and language that credits unused pool units rather than letting them expire at the annual reset.
If ServiceNow will not go below $0.08, trade for pool size and a shorter term rather than accepting a three-year lock, and benchmark the whole package against comparable accounts using what a good number looks like at your spend tier.
The clauses that decide your effective rate more than the rate does
Get a rate of $0.04 and you have won nothing if the surrounding clauses let ServiceNow reclaim it through consumption mechanics.
The five that matter most, in the order they cost buyers money: no-rollover annual resets, tenant-level pooling that sweeps sub-production, data quality as an undeclared multiplier, seat-scoped triggering rights, and the absence of auditable consumption reporting before a top-up invoice lands.
Each of these is a clause you insert, not a fact you accept.
No-rollover is the quiet one. Teiva Systems documents that each contract year stands alone, so unused assists evaporate at reset. That destroys the bank-surplus strategy and, more importantly, it makes over-forecasting a one-way transfer to ServiceNow.
If you commit to 2 million assists and burn 1.3 million, you funded 700,000 units of nothing. Insert either rollover of unused units into the following year (capped at 25 percent is a realistic landing zone) or a true-down right at each anniversary.
ServiceNow will resist rollover harder than it resists a lower rate, which tells you exactly how much the no-rollover clause is worth to them.
The pooling clause compounds it. Assists are pooled at the account level and consumed across all instances including sub-production, per ServiceNow's own community documentation, and usage tracks on a rolling 365-day basis. Your dev, test, and UAT instances burn production budget.
During an agentic build, where a single nine-to-twenty-tool execution costs 150 assists, a testing cycle can consume six figures of units before anything reaches a user.
Demand a separate non-production allocation, or a written exclusion for sub-prod consumption, and price your overage exposure past the bundled pool on the assumption you will not get it.
Data quality is the multiplier nobody prices. ServiceNow's own community post states plainly that an agent acting on a fractured CMDB retries, loops, and escalates, and that each step consumes assists, while a low Article Quality Index pushes AI Search into repeated lookups for the same answer.
That means your effective per-answer cost is a function of a data estate ServiceNow does not warrant. Do not accept that risk unpriced: tie a portion of the committed pool to remediation milestones, or negotiate a mid-term re-forecast right at month nine once real burn data exists.
Finally, entitlements are seat-scoped. The Starter Pack is scoped to 25 fulfiller users, so volume alone does not grant the right to trigger. And no top-up should be billable without instance-level, skill-level consumption reporting you can audit, with a 30-day dispute window in writing.
Evidence base: what the benchmarks and the packaging record show
Raymond James documents the published rate card at $0.20 per assist; advisory benchmarks put negotiated overage at $0.02 to $0.10, with Assist Packs at $50 to $150 per fulfiller per month.
A Starter Pack buyer who burns 150,000 assists in six months receives 500,000 more free, nominally worth $100,000 at card rate.
The record is consistent across independent sources.
ServiceNow's community confirms the ITSM Starter Pack ratio (25 Pro Plus users at 6,000 assists each) and the 500,000-assist bonus, while the CSM variant runs 250,000 assists for the same 25 fulfillers, proof the per-user ratio moves by product line.
Packaging is in motion: the April 2026 collapse of five legacy tiers into three, legacy end of sale in July 2026, and the June 2026 agentic schedule pricing executions at 25, 50, and 150 assists by tool count. Three patterns recur.
First, the rate falls fastest for buyers with a credible walk-away on scope, meaning you will drop Now Assist from two of five workflows rather than pay $0.20.
Second, the largest concessions arrive when ServiceNow wants a reference logo or a quarter-end adoption number, which is why the free bonus exists at all.
Third, and this is the pattern most buyers miss, those who settle the unit rate before debating the forecast land materially closer to $0.02 than to $0.10, because the volume conversation gives ServiceNow the cover it needs to hold the rate.
Benchmark the unit against your overall spend tier before the forecast discussion opens.
- 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
Your first five moves
- Pull 90 days of actual assist consumption by action type before you take a single call, broken out by skill (summarization at 1 assist, app creation at 20) and by instance, because sub-production testing draws from the same tenant-level pool and will otherwise show up as unexplained burn in your first true-up.
- Compute the Starter Pack implied rate at 650,000 units and open there, since the 150,000-assist bundle plus the 500,000-assist adoption bonus is ServiceNow's own published proof that it will transact at roughly 23 percent of the $0.20 card rate, which puts your opening counter at $0.02 to $0.03 and forces the rep to argue against his own promotion.
- Demand the top-up rate in writing inside the quote, locked for the full term, not in an order form footnote or a rate card reference, and require that the number survive co-terming, product additions, and the April 2026 tier migration; an unpriced overage clause is a blank check the vendor fills in at renewal.
- Name the workflows approved for agentic execution and cap everything else, because the 25/50/150 tiering means a nine-tool workflow costs $30 per run at card rate, and an uncapped agentic footprint quietly converts a unit negotiation into an unbounded one. Tie this to the burn-down mechanics covered in our Now Assist consumption and overage analysis.
- Set the walk-away as a dollar ceiling on total AI spend, not a rate, for example $400,000 across three years, so a good unit price cannot be undone by volume drift.
Frequently asked questions
What is ServiceNow's official overage rate for Now Assist?
ServiceNow's published rate card prices overage at $0.20 per assist, as documented by Raymond James in June 2026. However, ServiceNow does not publish a universal per-assist dollar rate in its standard price book, so $0.20 functions as an anchor rather than a contractual floor.
The rate that binds you is the one written into your quote.
What is a realistic negotiated top-up rate per assist?
Advisory benchmarks place negotiated overage rates between $0.02 and $0.10 per additional task, against a $0.20 card rate. Buyers with meaningful spend, multi-year commitments, or credible willingness to limit AI scope routinely land in the $0.03 to $0.05 range.
Anything at or near $0.20 means the rate was never negotiated.
How does the AI Starter Pack help me benchmark the unit rate?
The Starter Pack is the only public volume-to-package reference ServiceNow has documented: 25 Pro Plus users at 6,000 assists each for 150,000 total, plus a free 500,000-assist bonus if consumed within six months.
Dividing the pack price by 650,000 rather than 150,000 produces an implied rate roughly 77 percent below sticker. That number is ServiceNow's own arithmetic, which makes it hard for the rep to dismiss.
How many assists does an agentic workflow consume?
Per ServiceNow's June 2026 guidance, an agentic execution using zero to four tools consumes 25 assists, five to eight tools consumes 50, and nine to twenty tools consumes 150. At the $0.20 card rate that is $5.00, $10.00 and $30.00 per execution respectively.
At a negotiated $0.03 rate the same executions cost $0.75, $1.50 and $4.50.
Do unused assists roll over to the next contract year?
No. Assist entitlements reset annually and are tracked on a rolling 365-day basis, so unused volume from year one cannot cover overage in year two.
This makes over-forecasting expensive and removes the ability to bank surplus, which is why the unit top-up rate matters more than the size of the committed pool.
Does testing in a sub-production instance consume assists?
Yes. Assists are pooled at the account level and shared across all instances, including development, test and other sub-production environments.
Organizations running significant AI development work should either forecast that consumption explicitly or negotiate a separate non-production allowance, because otherwise development activity burns the same budget as production.
What changed on April 9, 2026 and why does it matter to the overage rate?
ServiceNow retired five legacy tiers (Standard, Pro, Pro Plus, Enterprise, Enterprise Plus) and replaced them with three AI-native tiers: Foundation, Advanced and Prime, each carrying a bundled assist pool. Legacy SKUs reached end of sale on July 1, 2026 and cannot be reinstated.
Because ServiceNow does not publish tier allocation values, buyers cannot derive an implied rate from the tier and must extract the top-up rate as a separately negotiated contractual number.
