The SaaS bill nobody owns, and how renewals can shrink it

Zylo, a vendor of SaaS management software, analyzed more than 40 million SaaS licenses and $75 billion in spend under management for its 2026 index. Its headline finding is that organizations leave an average of 36% of their SaaS licenses unused.¹ That is an average across the dataset, so it says nothing certain about any one company, and it counts licenses rather than dollars. Even so, it is a sensible place to begin a conversation about a line item that few finance teams can explain in full.

For a CFO the 36% matters less than the fact that nobody is accountable for it. In the same dataset, business units control 81% of SaaS spend and IT directly manages 15%.¹ A bill that many people buy and few people police drifts upward, and the drift stays invisible until a renewal arrives or a project is cancelled to pay for it.

Who actually buys the software

The ownership split explains most of what follows. When business units hold 81% of spend and IT 15%, the person who signs up for a tool is rarely the person who sees the invoice, and almost never the person who checks whether the licenses are used.¹ Larger firms feel this most. In Zylo’s data, large enterprises add an average of 21 applications per month.¹

Spend per head gives a sense of the size of the problem. Zylo’s 2026 index puts median SaaS spend per employee at $9,455.¹ The 2025 edition reported an average of $4,830 per employee, up 21.9% year over year.² The two editions rest on different datasets, and one figure is a median while the other is an average, so they cannot be set side by side as a trend.

Shadow AI enters through the expense report

The newest layer of the bill is the least visible. Zylo’s 2026 index reports that expense-based SaaS spend increased 267% year over year, with ChatGPT now the most expensed application, and it describes shadow AI as accelerating through expense-based purchasing.¹ A subscription paid with a company card tends to skip the purchase order process, which is why it often has no renewal date, no named owner and no usage review attached to it.

AI-native applications as a category are growing quickly. Zylo reports their spend up 393% year over year in organizations with more than 10,000 employees and up 108% overall.¹ Pricing is shifting too. In a survey of 218 IT leaders, 78% reported unexpected charges in the last 12 months tied to consumption-based or AI pricing models, and 61% said they were forced to cut projects because of unplanned SaaS cost increases.¹ Both percentages come from that survey of IT leaders, so they record experience as reported, not charges audited against invoices.

How renewal workflows and usage data cut the bill

The mechanics are simple. Usage data says which licenses are idle. A renewal workflow gives a named person the job of acting on that fact before the contract rolls over. Zylo’s 2025 edition recommends managing licenses and renewals programmatically.² Read plainly, that means the renewal date triggers a usage check, and the check decides the quantity.

Two design choices matter more than the tooling. The first is one owner per application, a person with a name rather than a department. The second is a fixed lead time. A notice period missed by a week turns a decision into a year of paying for the old quantity, and no amount of usage data repairs that.

Shadow AI needs the same loop with a different entry point. Expense lines and card statements have to be read as a source of applications, because that is where the fastest-growing subscriptions appear. Once an application is on the list it gets an owner and a date like any other. Where the tool is priced by consumption, there are no seats to remove, so the control becomes a spending cap and a monthly forecast instead of a seat count.

What it could be worth

Zylo’s 2025 edition reported that organizations in its dataset were wasting an average of $21M annually on unused SaaS licenses, a 14.2% increase year over year.² That dollar figure belongs to the 2025 dataset of over 40 million licenses and $40 billion in spend, and it should not be added to or compared directly with the 2026 percentages.

What the data does not say

Zylo sells SaaS management software, so it has a commercial interest in the size of this problem. The figures quoted here come from its press releases and are averages across its dataset. The 2026 release does not state how the unused share is distributed, so your own number may sit well above or well below 36%.

Unused is not the same as unneeded. Seats held for seasonal staff, auditors or business continuity look idle in a login report, and reclaiming them can create friction that costs more than the saving. Consumption pricing adds another uncertainty, because the survey above suggests many IT leaders are caught out by charges they did not forecast.¹ A seat count that falls while consumption charges rise is not a saving.

Finally, the 2026 release carries no figures on renewal price increases or on the share of shadow IT in total spend, so claims about those need another source. We looked for independent analyst figures on SaaS waste and could not verify any, so none are used here.

What to do next

1. Build one list of applications from contracts, card statements and expense lines, and put a named owner against each line.
2. Put every renewal date and notice deadline on a single calendar that finance maintains.
3. Pull per-seat activity before each renewal and set the seat count from it, not from last year’s count.
4. Route AI subscriptions bought on cards into the same list, and set a spending cap on any tool priced by consumption.
5. Compare your spend per employee with reference ranges in our SaaS spend benchmark, then track your unused share each quarter.

If you want a first estimate of where AI could cut cost in your own workflows, the free pre-audit is a short questionnaire that returns one.