Now Assist pricing and licensing: the buyer side guide

Section 01What Now Assist pricing means in 2026

Now Assist pricing changed shape in April 2026. AI is no longer a separate product you choose to buy; it is bundled into every tier of the new commercial model, Foundation, Advanced and Prime. What you pay for is not access to AI but the work it does, and that work is metered in a unit called an assist. Each tier includes a committed pool of assists, and when that pool is exhausted, overage triggers top up charges. This single shift, from a fixed add on to a metered consumption line, is the reason Now Assist pricing is both easier to start using and easier to underestimate than the AI products that preceded it.

For a buyer, the consequence is that the AI conversation is now a consumption conversation. The questions that decide cost are how many assists your workflows will consume, how that consumption is weighted across simple and agentic actions, what the overage rate is when you exceed the pool, and what flexibility you have if real usage diverges from the forecast. None of those questions is answered by the headline tier price, and all of them are negotiable.

This guide is the buyer side answer to Now Assist pricing, grounded in benchmark data from real enterprise renewals where we have sat buyer side in hundreds of enterprise software negotiations. It connects to the broader ServiceNow renewal pillar, and to the service we run alongside client teams, our ServiceNow renewal negotiation advisory. One scope note: this is commercial advisory guidance, not legal advice, and final contract language should be reviewed by counsel.

It helps to name why metered pricing feels different from the licensing buyers are used to. A traditional license is a known, fixed cost: you buy a number of fulfillers and the bill does not move until the next renewal. A metered line is a variable cost: it moves with usage, which means it can be lower than a fixed equivalent when usage is modest and considerably higher when usage grows. That variability is not inherently bad, and for many estates it is genuinely cheaper than a large fixed AI purchase would have been. The danger is that variability is harder to budget and harder to control, and a buyer who signs a metered line without guardrails has accepted an open ended cost that scales with their own success in adopting the tool.

The whole of this guide is therefore about converting an open ended cost into a managed one. That is done not by avoiding consumption pricing, which is now unavoidable, but by sizing the commitment honestly, fixing the overage rate, and securing the flexibility to adjust as real usage becomes known. Each of the sections below addresses one part of that conversion.

Section 02How assists are metered

An assist is the unit that meters Now Assist work. Conceptually, each AI action draws a quantity of assists from your committed pool, and the quantity scales with the complexity of the action. A simple generative request, summarising a record or drafting a short reply, consumes few assists. A multi step action that retrieves data, reasons across it and writes back to several records consumes more. The pool is depleted over the term, and the rate of depletion is a function of how much AI work your users and workflows actually do.

The practical difficulty for buyers is that assist consumption is hard to forecast from first principles, because it depends on adoption, workflow design and the mix of action types, none of which is fully known before deployment. This is precisely the gap the account team forecast fills, and it is precisely where buyers overcommit. A forecast built to justify a tier is not a consumption model, and the two should never be confused. We size assist commitments from a workflow level model that maps each automated workflow to an action type and an expected volume, then stress tests the total against optimistic and conservative adoption.

Understanding the metering matters because it converts a vague worry about AI cost into a specific negotiating position. A buyer who can describe consumption at the workflow level can argue for a committed volume on evidence, and can spot when a proposed commitment is padded. The detailed modelling sits in our Now Assist consumption advisory.

A second reason to understand the metering is that it reframes adoption as a cost driver, which it usually is not in fixed licensing. Under a per seat model, encouraging more users to adopt a feature has no marginal cost; the seats are already paid for. Under a metered model, every additional unit of AI work draws from the pool, so a successful adoption programme accelerates consumption and, beyond the commitment, triggers overage. This does not mean adoption should be discouraged, which would waste the platform. It means adoption and consumption must be planned together, so that the commitment and the overage terms anticipate the success the organisation is trying to create rather than being surprised by it.

Section 03Agentic actions and consumption weighting

The most important nuance in Now Assist pricing is that not all AI actions cost the same. Large agentic actions, those where the system chains multiple steps, calls tools and acts with a degree of autonomy, consume materially more assists than simple generative requests. As agentic automation moves from pilot to production, the consumption weighting of those actions becomes the dominant driver of total assist usage, and an identical committed volume can prove generous or wholly inadequate depending on how agentic your roadmap is.

This creates a specific risk at the point of commitment. A buyer who sizes an assist pool against today's mostly generative usage, then deploys agentic workflows over the term, can exhaust the pool far earlier than the forecast implied and fall into overage at exactly the moment AI is delivering the most value. The defence is to document the consumption weighting of agentic actions in the agreement, model exposure across a realistic agentic adoption curve, and secure flexibility terms that absorb the divergence.

It is worth being concrete about why agentic actions consume so much more. A simple generative request is a single inference: a prompt goes in, a response comes back. An agentic action is a sequence: the system plans, retrieves context, calls one or more tools, evaluates results, and often loops before completing, with each step drawing on the model. A single agentic resolution of a service request can therefore represent many times the assist draw of a one shot summary. The implication for sizing is that counting AI actions tells you very little; what matters is the weighted volume, where agentic actions are weighted up sharply. A consumption model that counts actions without weighting them will understate true consumption precisely for the workloads an organisation is most likely to expand.

This is also why the same committed volume can be a bargain for one enterprise and a trap for another. Two organisations might project an identical number of AI actions per month, but if one is largely generative and the other is heavily agentic, their real assist consumption can differ by a wide margin. The committed volume that protects the first leaves the second in overage within months. The only defence is a weighted model built from each organisation's own workflow mix, which is exactly the work the vendor forecast does not do.

The tier you migrate to interacts with this directly, because agentic capability and assist allocation differ across Foundation, Advanced and Prime. Our spoke on the ServiceNow Advanced tier works through where agentic workloads land in the new model.

Section 04AI bundled across Foundation, Advanced and Prime

In the 2026 model, the five legacy tiers of Standard, Pro, Pro Plus, Enterprise and Enterprise Plus were replaced by Foundation, Advanced and Prime, and AI is bundled into all three. The bundling is genuinely useful in that it removes a separate purchase decision, but it also changes the negotiation. You can no longer defer AI as a future line item; instead, the assist allocation that comes with your tier, and the overage terms beyond it, are part of the tier negotiation itself.

The buyer side implication is that tier selection and assist sizing must be negotiated together, not in sequence. A higher tier that bundles a larger assist allocation may be cheaper in total than a lower tier plus heavy overage, or it may be an upsell dressed as efficiency, depending on your real consumption. The only way to know is to model consumption first and let the model drive the tier conversation, rather than accepting a tier and discovering the assist economics afterward.

This is also where the tier migration of a renewal collides with Now Assist pricing. A forced move up the tier ladder may be justified to the buyer partly on the AI allocation it bundles, so the value of that allocation has to be assessed honestly against modelled consumption. The mechanics of one common migration path are in our spoke on the ServiceNow Pro Plus to Advanced migration.

There is a subtle trap in how the bundled allocation is presented. Because the allocation arrives with the tier at no separately stated price, it is easy to treat it as free, and therefore to value any overage avoided by a larger allocation as pure saving. But the allocation is not free; its cost is embedded in the tier price, and a larger allocation that justifies a higher tier is being paid for whether it is used or not. The correct comparison is always between the total cost of a lower tier plus expected overage and the total cost of a higher tier with its bundled allocation, computed against the same modelled consumption. Run honestly, that comparison sometimes favours the lower tier with overage, and sometimes the higher tier, but it is never settled by the headline allocation number alone.

Section 05Sizing your assist commitment

Sizing the committed assist volume is the central act of Now Assist pricing, and it follows a short discipline.

The principle throughout is to commit to what you can defend and protect the rest with terms. Prepaying a large assist pool against an optimistic forecast converts uncertainty into shelfware; committing conservatively and negotiating overage converts the same uncertainty into a managed cost.

A reasonable objection is that conservative commitment risks paying overage rates that are higher than the committed rate, so where is the gain? The gain is in who carries the uncertainty. A large upfront commitment makes the buyer carry it, because unused assists are paid for and rarely refunded. A conservative commitment with a negotiated overage rate moves the uncertainty into a variable cost that is only incurred if the usage materialises, and that, crucially, only materialises alongside the value the usage created. Paying a slightly higher unit rate on assists you actually used, in workflows that are demonstrably delivering, is a far better position than having prepaid for assists that adoption never reached. The arithmetic favours conservatism in every scenario except the one where adoption exactly matches an aggressive forecast, which is the scenario that almost never occurs.

Section 06Overage and top up charges

When consumption exceeds the committed assist pool, overage triggers top up charges. Overage is not inherently bad; it is the mechanism that lets a conservative commitment scale with real usage. What makes overage dangerous is an unfavourable rate negotiated at the wrong time. An overage rate agreed at signature, while you still have leverage and alternatives, is a managed cost. An overage rate discovered mid term, when the pool is exhausted and the workflows depend on it, is whatever the vendor decides it is.

Three protections turn overage from a risk into a tool. First, the overage rate itself, fixed in the agreement at signature. Second, rollover or true forward treatment, so unused assists in one period offset consumption in another rather than expiring. Third, a midterm resize right, so that if real consumption settles materially above or below the commitment, the commitment can be adjusted to match rather than locking in a number set before deployment. Based on benchmark observations, enterprises that secure all three see materially lower effective AI costs than those that accept printed consumption terms, even when both start from a similar committed volume.

It is worth understanding why the timing of the overage conversation matters so much more than its substance. At signature, the buyer has alternatives: a lower commitment, a different tier, a slower AI rollout, or simply walking the whole renewal. Those alternatives give the overage rate a ceiling, because an unreasonable rate can be met by declining to depend on overage at all. Mid term, once workflows are live and the pool is exhausted, those alternatives have evaporated. The workflows cannot be switched off without operational cost, and the only way to keep them running is to consume overage at whatever rate the contract left open. This is the same leverage dynamic that governs the whole renewal, concentrated into a single term, and it is why the overage rate is one of the few numbers that should never be deferred.

Rollover and true forward treatment deserve a similar word, because their value is easy to underestimate. Without rollover, any gap between a conservative commitment and actual consumption is simply lost: assists paid for and not used expire at the period boundary, which quietly raises the effective price of every assist the organisation did consume. With rollover, that same gap becomes a buffer against future peaks, so the conservative commitment is protected on both sides. For an organisation whose AI adoption is uneven across the year, which is most of them, rollover can be worth more than a modest improvement in the headline rate, and it costs the vendor little to grant, which makes it one of the more winnable terms in the negotiation.

Section 07Negotiating Now Assist at renewal

At renewal, Now Assist pricing collapses into a small set of negotiable terms. Treat them as an agenda.

1. Committed assist volume

   Sized from your workflow level model, committed toward the conservative end. The volume is the largest single number in the AI line, and the easiest to overcommit.

2. Overage rate

   Fixed at signature, never left open. On a multi year term, a negotiated overage rate is often worth more than an additional point of discount elsewhere.

3. Rollover and true forward

   Unused assists should offset future consumption rather than expire. Expiry quietly inflates the effective price of every assist you committed to and did not use.

4. Midterm resize right

   The right to adjust the commitment if real consumption diverges from the model protects you in both directions and keeps the agreement honest over the term.

5. Agentic weighting transparency

   The consumption weighting of agentic actions documented in the agreement, so the cost of your roadmap