Unattended robot utilization ran 40 to 65 percent of licensed capacity once the schedules and queues were actually audited
Automation estates age quickly. Process changes retire bots faster than license administrators notice, and the renewal quietly reprices the difference.
Prepared by Redress Compliance · August 19, 2026 · UiPath agreements. 12 to 18 agreements reviewed, 2024 to 2025.
Executive summary
Unattended robot utilization ran 40 to 65 percent of licensed capacity once schedules and queues were audited against the orchestration logs rather than the order form.
Orphaned licenses from retired automations accounted for 10 to 20 percent of renewal value before any cleanup, attached to processes that no longer exist.
Estates that priced a commodity workflow alternative settled 15 to 25 percent better. Partial displacement is the threat that actually prices.
Agentic capacity is the new default uplift in quotes. Treat those units as a separate negotiation with their own measured baseline.
How does the licensing actually work?
Through platform editions and units covering robots, development seats and the AI capabilities. The three license types price very differently from one another.
Three things you buy, one that dominates
- Unattended robots: the expensive core, licensed per concurrent runtime.
- Attended robots: per user licenses for human triggered automation.
- Platform and AI units: consumption units covering orchestration, document understanding and agentic features.
Concurrent runtime is not headcount
The unattended line is bought as concurrency, which means the question is how many robots run at the same moment rather than how many automations exist. Those two numbers diverge quickly. The structure is published on the pricing page with the mechanics in the licensing documentation.
How do you audit the estate before renewal?
From the orchestration logs, at least 90 days out. Scheduled runtime against licensed capacity, per automation. The gap is usually wide and always negotiable.
| What the audit surfaces | Where it hides | What it is worth | How to fix it |
|---|---|---|---|
| Idle capacity | Robots scheduled far below licensed concurrency | The 40 to 65 percent utilization gap | License to measured runtime |
| Orphaned licenses | Robots attached to retired automations | 10 to 20 percent of renewal value | Reclaim before the quote |
| Studio sprawl | Development seats held by teams no longer building | A quiet per seat line | Reassign or drop at renewal |
| Unit drift | AI and platform units bundled in past renewals | Capacity never consumed | Reset to measured consumption |
Renew the measured number, not the licensed one
Renewing the licensed number instead of the measured one is the single most common overspend on this platform. The number on the order form is the number nobody checked.
Ninety days is the working minimum
The log pull, the per automation reconciliation and the reclaim cycle do not compress into a quarter end. Starting later means arriving with the vendor's count as the only count.
The UiPath negotiation guide
The measured capacity method, the alternative anchor, and the buyer side moves across an automation estate.
Get the brief →What 12 to 18 UiPath agreements showed
Across roughly 12 to 18 UiPath agreements Fredrik Filipsson reviewed between 2024 and 2025, licensed capacity outran deployed automation in almost every estate. Three patterns recur.
- Unattended robot utilization ran 40 to 65 percent of licensed capacity once schedules and queues were audited.
- Orphaned licenses from retired automations accounted for 10 to 20 percent of renewal value before cleanup.
- Estates that priced an alternative for commodity workflows settled 15 to 25 percent better on the renewal.
The platform pages sell transformation. Your orchestration logs sell reality, and only one of the two is admissible at the table.
- Percentile standing for your exact deal size and industry, from real closed transactions
- Every risky clause flagged with the exact quote, the page, and the replacement language to send back
- A negotiation playbook, talking points, and a two page executive brief on day one
Which levers actually move the quote?
Measured utilization, a costed alternative for commodity flows, and per automation return math. Nothing else reliably moved paper in the reviewed file.
Four levers and when each one works
- Utilization audit, from the logs 90 days out: cuts 10 to 20 percent of orphaned value.
- A costed alternative for commodity workflows: 15 to 25 percent better settlement.
- Per automation return file, cost against delivered hours: disciplines the expansion pitch.
- Term traded for a written uplift ceiling: protects against unit repricing.
Why the adjacent platform anchor is the credible one
The commodity workflow alternative usually ships inside paper the enterprise already holds, so the marginal cost argument is believable even when the complex estate stays put. Partial displacement is the threat that prices.
Keep the commodity flows portable
Governance simplicity is real, and it is cheaper to buy with architecture than with a monopoly renewal. The wider framework sits in the automation procurement framework.
How should the agentic units be handled?
As a separate negotiation with its own measured baseline, because bundled agentic capacity is the new default uplift in quotes.
Three rules that hold on the AI line
- Pilot before committing: measure document understanding and agent consumption on real volumes first.
- Reject default bundles: units belong on the order form only at measured quantities.
- Reprice annually: unit pricing moves fast, and long commitments at today's rates favour the vendor.
Your renewal is where the growth story is priced
Agentic automation is the vendor's growth narrative, as its investor communications make plain, and the renewal is where that narrative meets your order form. The clause work sits in the AI contract playbook.
Where the common advice on automation is wrong
The standard advice says consolidate all automation on one vendor to maximize discount tiers and governance simplicity. We disagree.
Single vendor estates paid more per delivered hour
Across the agreements reviewed, single vendor estates paid more per delivered automation hour than mixed estates, because the consolidation discount never offset the leverage lost when the alternative stopped being live.
The buyer side move is to keep commodity workflows portable, license unattended capacity to measured utilization, and negotiate the AI units separately at pilot proven volumes. The product framing is on the platform pages, and the estate picture in the spend health check.
What the agreements measured, 2024 to 2025
Two cuts of the review file, both recoverable from your own logs.
Once schedules and queues were audited from the orchestration logs rather than read from the order form.
Attached to automations retired by process or system change, and still on the renewal until somebody reclaimed them.
Neither number requires the vendor to concede anything. Both require somebody to pull the logs before the quote lands.
Your first five moves
- Pull the orchestration logs at least 90 days before renewal, because the reconciliation and the reclaim cycle do not compress into a quarter end.
- Measure scheduled runtime against licensed concurrency per automation, which showed utilization running 40 to 65 percent across the reviewed estates.
- Reclaim the licenses attached to retired automations, since orphaned capacity accounted for 10 to 20 percent of renewal value before any cleanup.
- Cost a commodity workflow alternative and keep it live, because estates that did settled 15 to 25 percent better on the renewal.
- Negotiate the AI units separately at pilot proven volumes. The vendor management practice runs the utilization audit before the quote arrives.
Frequently asked questions
What drives the bill most?
Unattended robot capacity, licensed per concurrent runtime. It is the expensive core, and concurrency is not the same number as the count of automations you have built.
How utilized is a typical estate?
Between 40 and 65 percent of licensed capacity once schedules and queues were audited from the orchestration logs rather than read from the order form.
What are orphaned licenses?
Capacity attached to automations retired by process or system change. They accounted for 10 to 20 percent of renewal value before anybody reclaimed them.
When should the audit start?
At least 90 days before renewal. The log pull, the per automation reconciliation and the reclaim cycle do not fit inside a quarter end.
Does a competitive alternative help?
Materially. Estates that costed an alternative for commodity workflows settled 15 to 25 percent better, because partial displacement is a credible threat.
Why is that anchor credible?
Because the commodity workflow alternative usually ships inside paper the enterprise already holds, so the marginal cost argument stands up even if the complex estate stays.
How should the AI units be treated?
As a separate negotiation with its own measured baseline. Bundled agentic capacity is the new default uplift, and it belongs on the order form only at measured quantities.
Should the AI commitment be long?
No. Unit pricing is moving fast, so a long commitment at today's rates favours the vendor. Reprice annually and pilot on real volumes first.
Is single vendor consolidation cheaper?
Not in the reviewed file. Single vendor estates paid more per delivered automation hour, because the discount never offset the leverage lost when the alternative went away.
What is the most common overspend?
Renewing the licensed number instead of the measured one. The figure on the order form is the figure nobody checked against the logs.
