Here is the short version. The AI add-on is now a large enough line to deserve its own negotiation, and the three leading ITSM platforms have chosen three different pricing models, which makes a like-for-like rate comparison impossible. A per-user figure, a consumption meter and a hybrid cannot be reduced to one number without distorting the picture. What you can compare is landed cost: model each vendor's AI against your actual volumes and use cases, then negotiate the model, not just the rate. This article sits in our ITSM total cost of ownership calculators toolkit.
Pricing AI across two or more ITSM vendors? Book a contract negotiation review and we will model the landed cost per vendor before you commit.
How are ITSM AI add-ons priced in 2026?
By three different mechanisms. ServiceNow Now Assist is priced on consumption, metered in assists, and from April 2026 AI capabilities including Now Assist, Moveworks, Workflow Data Fabric and the AI Control Tower are bundled into the higher tiers, with legacy SKUs reaching end of sale on 1 July 2026. Atlassian Rovo is a per-user add-on priced at about $20 per user per month, sitting on top of Jira Service Management. Freshservice Freddy combines per-agent and consumption-based elements. The differences are the whole story: each model rewards a different usage profile, so the cheapest vendor for one buyer is the most expensive for another.
Why is comparing these prices so hard?
Because the units are not the same. A per-user add-on charges for access regardless of use, which favours light users and punishes heavy ones. A consumption meter charges for use regardless of headcount, which favours a small team automating a lot and punishes broad low-intensity rollout. A hybrid splits the difference. Reduce these to a single per-user rate and you will reach the wrong answer, because the same headline figure produces very different bills depending on how your team actually works. The only honest benchmark is the landed cost for your volumes.
How do the 2026 models compare?
The table compares the pricing model and what it rewards, not a single price, because the published mechanisms differ and any single dollar figure would mislead. Confirm current rates against your order form.
| Add-on | 2026 pricing model | What it rewards | Key buyer-side watch-out |
|---|---|---|---|
| ServiceNow Now Assist | Consumption, metered in assists; bundled into higher tiers from April 2026 | Concentrated, high-value automation by a smaller team | Volume forecast risk and a possible tier uplift to access AI. |
| Atlassian Rovo | Per-user add-on, about $20 per user per month | Broad access where most users engage AI | Paying per seat for users who rarely invoke it. |
| Freshservice Freddy | Mixed per-agent and consumption | Mid-size teams with steady, predictable use | Two cost drivers to track, not one. |
What does this mean for a consumption model specifically?
If your platform uses a consumption meter, the bill is a volume forecast you have to build, not a per-seat line you can read off a price list. That forecast is where the money is made or lost, which is why we model assists bottom-up from real ticket data before sizing any commitment. The full method for ServiceNow specifically is in the Now Assist consumption estimator, and the same discipline applies to any metered AI: size the base to evidence, cap the overage, and re-price once usage is known.
What should you negotiate on an AI add-on?
- Separate the AI line from the subscription so its growth is visible and cannot hide inside the renewal total.
- Right-size any committed volume to your modelled usage, starting below the vendor's proposal.
- Cap the overage rate so unexpected demand does not price at a premium.
- Take a short AI term so you can re-price once you hold real usage data.
- Resist a tier uplift you do not otherwise need just to access AI, and price that uplift as part of the AI cost if it is unavoidable.
Our gated 2026 ITSM Negotiation Benchmark Report includes the AI add-on model comparison and the landed-cost worksheet behind this article.
What to do next
Do not compare AI add-ons on the headline rate. Build a landed-cost model per vendor against your real volumes, then negotiate the model: separate the line, right-size the commitment, cap the overage and keep the term short. We model and negotiate ITSM AI add-ons on fixed fee or gainshare through our contract negotiation service, and since 2019 the firm has negotiated more than $420M in ITSM contract value at a 30% average reduction across 500-plus engagements.
Frequently asked questions
- How are ITSM AI add-ons priced in 2026?
- Differently by vendor. Now Assist is consumption-metered in assists and bundled into higher tiers from April 2026. Rovo is a per-user add-on at about $20 per user per month. Freddy mixes per-agent and consumption. Compare on landed cost, not a single rate.
- Why is comparing AI add-on prices across vendors so difficult?
- Because the pricing units differ. A per-user add-on, a consumption meter and a hybrid are not comparable on one rate. Model the landed cost for your volumes and use cases.
- What should I negotiate on an ITSM AI add-on?
- Separate the AI line, right-size committed volume, cap overage, take a short term, and resist a tier uplift you do not otherwise need just to access AI.
Book a contract negotiation review.
We model AI add-on landed cost per vendor and negotiate the model, not just the rate. Fixed fee or gainshare.
Book a contract negotiation review →