Now Advisory · Buyer side guide · 2026 edition
Now Assist ROI: a buyer side guide
Now Assist ROI is only real if the assist economics underneath it hold. This guide shows how to build a return case from metered consumption and genuine workflow value rather than vendor adoption hopes, with benchmark data from real enterprise renewals.
Section 01What Now Assist ROI really measures
Now Assist ROI measures whether the value a workflow gains from AI exceeds the metered cost of the assists it consumes. In the 2026 model, AI is bundled into every tier and metered in assists, so the return is no longer a flat licence question; it is a consumption question. A workflow can show a strong return at pilot volumes and a poor one at production volumes if the assist draw scales faster than the value, which is why ROI has to be modelled against consumption, not asserted at a headline.
The buyer side distinction is between value delivered and assists consumed. Vendor ROI material tends to lead with value, the hours saved, the deflection achieved, while leaving the consumption side soft. A defensible Now Assist ROI case holds both columns to the same standard: the value is estimated conservatively and the assist cost is modelled from realistic usage, including the heavier draw of agentic actions.
This guide sits under the Now Assist pricing pillar and pairs with our Now Assist consumption advisory. It treats ROI as a number you defend, not a number you receive.
Section 02The assist economics underneath
Every Now Assist ROI rests on three economic facts. AI is bundled into the tier, so the committed assist pool is part of what you already pay; assists are metered, so consumption beyond the pool is billable; and large agentic actions consume materially more assists than simple generative requests. The return on any workflow is therefore the value it produces divided by the assists it draws, and the second number is the one most ROI cases get wrong.
Because the pool is committed, the marginal economics shift at the pool boundary. Inside the pool, an additional assist is effectively prepaid; beyond it, the same assist is charged at the overage rate as a top up. An ROI case that ignores where consumption sits relative to the pool can flatter the return by treating all assists as free or penalise it by treating all as overage. The honest model tracks consumption against the pool and prices each band correctly.
The assist draw of a workflow at production scale is frequently several times its draw at pilot, driven by agentic adoption. ROI cases built on pilot consumption routinely overstate the return once the same workflow runs at full volume.
Section 03Where ROI cases break
Now Assist ROI cases break in recognisable ways. The most common is pilot extrapolation, where a favourable pilot return is projected linearly to production while consumption scales faster than value. The second is uncounted agentic weight, where the model treats every AI action as a single assist when agentic resolutions draw many. The third is ignored overage, where the case assumes all consumption sits inside the pool when production volumes push it beyond, into top up charges.
A fourth break is value optimism: counting the full theoretical time saving rather than the realised one, when adoption, accuracy and rework all erode the headline. None of these is a reason to avoid Now Assist; they are reasons to model it honestly, so the return that justifies the spend is one that survives contact with production. Our Now Assist overage guide details the consumption side of this exposure.
A break to watch for specifically is the ROI that depends on a pool sized to vendor adoption hopes; if the pool is too large, the prepaid waste quietly lowers the real return.
Section 04Building the return model
A defensible model has two sides built to the same rigour. On the value side, estimate the realised benefit per workflow conservatively: hours saved at a loaded rate, deflection at an observed rate, quality gains where they can be evidenced, all discounted for adoption ramp. On the cost side, model the assist draw per workflow at production volume, weighting agentic actions, and price consumption in two bands, inside the committed pool and beyond it at the overage rate.
The return is then value minus modelled assist cost, expressed as a range across conservative and optimistic adoption rather than a single point. A range is more defensible than a number because it shows finance the floor and the ceiling, and it tells the negotiation team how sensitive the return is to the assist rate and pool size. The modelling method connects to our Now Assist consumption advisory.
Build the ROI as a range, not a number. A single ROI figure invites a single objection; a range built from conservative and optimistic adoption survives scrutiny and shows exactly which assist terms matter most.
Section 05Weighting agentic actions
The weighting of agentic actions is the hinge of any honest Now Assist ROI. A simple generative request is roughly a single inference; an agentic action plans, retrieves context, calls tools, evaluates results and often loops, so one agentic resolution can represent many assists. An ROI model that treats both as equal will overstate the return wherever agentic workflows dominate, which is increasingly everywhere.
The buyer side practice is to estimate the assist weight of each workflow class from its actual behaviour, then apply that weight to forecast volumes. This is also why the documented weighting belongs in the contract: if the agreement leaves agentic weighting undefined, the ROI you modelled and the consumption you are billed for can diverge sharply. Our Now Assist licensing guide covers the contractual side.
Weighting is not a reason to avoid agentic workflows, which often carry the highest value; it is the reason to size their assist draw honestly so the high value is set against its true cost.
Section 06Protecting the ROI at renewal
An ROI that holds in the model can still erode in the contract if the commercial terms are left open. The return depends on the assist rate, the overage rate, the pool size and rollover treatment, and each is negotiable at renewal. A strong modelled return on weak terms is fragile: a mid term overage rate or an expiring pool can turn a positive case negative without the value ever changing.
Protecting the return means fixing the overage rate at signature, sizing the pool to modelled consumption, securing rollover so a conservative commitment is not penalised, and capping assist rate increases for the term. These are the terms that keep the modelled ROI real, and they are won at renewal while alternatives are live. The negotiation detail sits in our Now Assist negotiation guide.
Treat the ROI and the contract terms as a single artefact: the model tells you which terms matter most, and those become the priorities at the table.
Section 07The ROI model, line by line
A defensible Now Assist ROI reduces to a small set of evidenced lines, each modelled rather than asserted.
- Realised value per workflow
Conservative benefit estimates discounted for adoption ramp, not headline theoretical savings.
- Weighted assist draw
Production volume consumption per workflow, with agentic actions weighted to their true assist cost.
- Two band consumption pricing
Assists priced inside the committed pool and beyond it at the overage rate, tracked against the pool boundary.
- Return as a range
Value minus modelled cost across conservative and optimistic adoption, not a single point.
- Term sensitivity
The lines showing how the return moves with assist rate, pool size and overage, marking the negotiation priorities.
Built this way, the ROI becomes both a budgeting tool and a negotiation map: it tells finance the range and tells the negotiation team exactly which terms protect it.
Section 08The Now Assist ROI checklist
Before presenting a Now Assist ROI internally or at renewal, confirm each item below from modelled evidence rather than pilot extrapolation.
- Value estimated conservatively and discounted for adoption ramp.
- Assist draw modelled at production volume, not pilot.
- Agentic actions weighted to their true assist cost.
- Consumption priced inside the pool and beyond it at the overage rate.
- Return expressed as a range across conservative and optimistic adoption.
- Overage rate, pool size, rollover and rate cap identified as the terms protecting the return.
If the return depends on pilot consumption or unweighted assists, it is not yet an ROI; it is a hope. The number you take to finance should be the one the consumption model defends.
FAQFrequently asked questions
What does Now Assist ROI measure?
Now Assist ROI measures whether the value a workflow gains from AI exceeds the metered cost of the assists it consumes. Because AI is bundled and metered in the 2026 model, the return is a consumption question: a workflow can show a strong return at pilot volumes and a weak one at production volumes if assist draw scales faster than value.
Why do Now Assist ROI cases overstate returns?
The common causes are pilot extrapolation projected linearly to production, treating agentic actions as single assists when they draw many, ignoring overage when production volumes exceed the pool, and counting theoretical rather than realised value. Modelling consumption honestly corrects all four.
How should agentic actions be weighted in ROI?
A simple generative request is roughly one inference, while an agentic action plans, retrieves, calls tools and loops, so it can draw many assists. Estimate the weight per workflow class from real behaviour and apply it to forecast volumes, and make sure the agentic weighting is documented in the contract.
How do we protect a Now Assist ROI at renewal?
Fix the overage rate at signature, size the pool to modelled consumption, secure rollover so a conservative commitment is not penalised, and cap assist rate increases for the term. These terms keep the modelled return real and are won at renewal while alternatives are live.
Are these figures official ServiceNow prices?
No. All ranges are typical negotiated figures based on benchmark observations across real enterprise renewals, used as internal leverage rather than published as official list prices.