Here is a number that should be more scandalous than it is. According to Zylo’s 2026 SaaS Management Index, the average organization wastes $19.8 million a year on unused software licenses — seats that were purchased, provisioned, budgeted, and then never opened by anyone.
The instinctive reaction is that this is a discipline problem: sloppy procurement, lazy IT, companies that need to get organized. That reading is comfortable and mostly wrong. The waste is not a failure of the system. It is what the system produces when it works correctly — and the 2026 data shows exactly why.
Sprawl stopped growing. Waste kept rising.
The most revealing statistic this year is not the size of the waste but its direction relative to the app count. The average company’s portfolio has settled at roughly 305 applications and largely stopped growing, yet the wasted spend climbed more than 14 percent year over year. Companies are not buying dramatically more tools. They are getting worse at using the ones they already have.
The mechanism is visible in the utilization rate: only about 49 percent of provisioned licenses are actually used. Roughly half of every software invoice buys nothing. And that ratio deteriorating while the app count holds steady tells you the problem has migrated. This is no longer a story about adoption running wild. It is a story about retirement — about the fact that organizations are structurally good at starting subscriptions and structurally terrible at ending them.
Why cancellation is the hard part by design
Think about the asymmetry in effort. Adopting a SaaS tool takes a corporate card and about four minutes. Nobody needs approval, nothing gets installed, and the vendor has spent enormous engineering effort making that path frictionless. Cancelling requires someone to notice the line item, determine who owns it, confirm nobody depends on it, find the contract, identify the renewal window, and act before it auto-renews. One of those is a click. The other is a project.
Renewal, meanwhile, is the default state. It requires no decision from anyone. A subscription that everyone forgot about behaves exactly like a subscription everyone loves — the invoice is identical. This is the core of why the waste persists through every governance initiative aimed at it: the business model is engineered so that inertia produces revenue. Not maliciously, necessarily. But a product priced per seat per month, sold with automatic renewal and self-service signup, is a product optimized to keep getting paid after it stops being useful.
That is why the sprawl statistics never improve despite a decade of executives promising to consolidate. Governance is a periodic human effort fighting a continuous automated process. The automated process wins on the years nobody runs the cleanup.
AI made the meter run
The genuinely new development in 2026 is that AI tools are breaking the one thing that made SaaS budgeting tractable: predictability. Per-seat pricing was wasteful but at least it was knowable — you could multiply seats by price and get next year’s bill. AI-native tools increasingly price by consumption, and the result is that 66.5 percent of IT leaders reported surprise charges from consumption-based AI pricing, with AI-native application spend growing 75.2 percent in a single year — the fastest-growing category on the invoice.
Combine those two facts and the exposure changes shape. Under per-seat pricing, an unused tool costs you a fixed, capped amount. Under consumption pricing, a tool nobody is monitoring can cost an unbounded amount, because the meter runs on usage rather than headcount — including automated usage that no employee ever initiates. The old waste was paying for something nobody opened. The new waste is paying for something running without anyone watching.
The security dimension is arriving on the same schedule. Torii’s 2026 benchmark found the average large enterprise operating 2,191 applications, with more than 61 percent not formally approved or overseen by IT. As the company’s chief executive Uri Haramati put it, “AI didn’t create shadow IT, but it dramatically increased its speed and blast radius.”
One caveat about all these numbers
Worth saying plainly, because almost no coverage of this topic does: nearly every statistic above comes from a company that sells software to fix the problem it is measuring. SaaS management platforms produce the SaaS waste research. That does not make the numbers wrong — they are internally consistent across several independent vendors, and the underlying dynamic is obviously real to anyone who has audited a corporate card statement. But the incentive runs in one direction, and a figure like “$19.8 million” is doing marketing work as well as analytical work. Treat the direction of these findings as reliable and the precise magnitudes as advocacy-adjacent.
The practical response does not require buying anything, which is convenient given who is publishing the research. It requires converting renewal from a default into a decision. Put every subscription on a calendar with a named owner and a date thirty days before its renewal, and make the owner affirmatively re-justify it rather than letting silence equal yes. That single inversion — opt-in renewal instead of opt-out cancellation — addresses more of the waste than any consolidation project, because it attacks the actual mechanism rather than the symptom. The tools were never the problem. The default was.
Related reading: Everyone’s Adopting AI. Almost Nobody’s Profiting. That’s the Real 2026 Story
