- Software license checkouts hide the real cost: about 51 percent of licenses are underutilized even though they look active.
- AI seat add-ons are billed for an entire user base on day one, while first-year adoption often reaches only 15 to 35 percent.
- Software renewals are priced off your existing baseline, not actual demand, so entering one without usage data compounds overspending every year.
When an engineer opens a CAD or CAE license at 8:30 one morning, the license manager marks it checked out — and it stays that way through the stand-up, lunch, a two-hour design review, and an afternoon spent mostly in email, until it closes past 5:30. Nine hours on the books, three of them real work. It is a pattern repeated across concurrent license pools in engineering organizations everywhere, and it is why IT budgets rarely fail from one bad purchase — they bleed quietly, through spending that is fully approved, documented, and invisible to the people held accountable for it.
The line items you can see are rarely the ones doing the damage.
Everyone remembers the iceberg. Almost nobody remembers the messages.
On the afternoon of April 14, 1912, the Titanic’s wireless room received a series of ice warnings from other ships crossing the North Atlantic. Several of them never made it to the bridge. The lookouts in the crow’s nest that night were working without binoculars, a detail historians have traced back to a last-minute crew change before the ship left Southampton. The iceberg was the event. The missing information was the cause.
Corporate IT budgets fail in much the same way. They rarely collapse because of one bad purchase that everybody can point to. They bleed quietly, through spending that is fully approved, properly documented, and almost completely invisible to the people held accountable for it.
The pressure has never been higher. Gartner projects global software spending will reach $1.44 trillion in 2026, up 15.1 percent year over year. Meanwhile Zylo’s 2026 SaaS Management Index found that application counts have flattened while spend still climbed 8 percent. Organizations are not buying more tools. They are paying more for the ones they already own.
Here are three costs that almost never show up as a line item, and what it takes to see them.
1. The Gap Between a License Checked Out and a License Actually Used
Most organizations already understand shelfware: the tool nobody opened, the pilot that stalled after the pilot. Vertice’s Q1 2026 benchmark data, drawn from more than $30 billion in processed software spend, puts outright shelfware at 15 percent of licenses and another 51 percent underutilized.
The 15 percent gets cleaned up. The 51 percent is where the real money sits, and it is far harder to find, because those licenses look busy.
This gets expensive fastest in engineering environments, where a single seat of high-end CAD, CAE, or subsurface software can run well into five figures a year. When an engineer opens that application at 8:30 in the morning, the license is checked out. It stays checked out through the stand-up, through lunch, through a two-hour design review, and through the afternoon spent mostly in email. Your license manager records one continuous checkout of just over nine hours. The actual work took about three.
Multiply that across a shared pool of concurrent licenses and the pattern turns costly in a very specific way. Someone who genuinely needs the tool gets denied, so the team requests more licenses. The pool grows. Utilization drops further. The budget expands to solve a shortage that was never really a shortage.
Checkout data cannot reveal any of this. Only activity-level metering can, because the real question is not whether the application is open. It is whether anyone is touching it.
2. The AI Premium You Are Paying for at Full Price
This cost is new, and it is moving faster than most budget cycles can track.
Every major vendor has now folded AI into its price list. Microsoft layered Copilot into Microsoft 365. Salesforce built Einstein and Agentforce into its higher CRM tiers. SAP is threading Joule across its cloud portfolio. Some of it arrives as an add-on with its own per-user fee. Some arrives as a forced tier migration, where the SKU you renewed for years quietly retires and the only replacement costs more.
The problem is not that AI features cost money. It is that they are almost always bought for the entire user base on day one, while adoption follows the same slow curve every enterprise rollout has ever followed. VendorBenchmark’s March 2026 analysis puts first-year adoption of AI seat add-ons somewhere between 15 and 35 percent. Deloitte’s 2026 State of Generative AI in the Enterprise study found that 72 percent of enterprise AI projects exceed their original budget by at least 30 percent.
Then there is the meter. Consumption pricing, billed in tokens, credits, conversations, or resolutions, does not sit neatly inside an annual budget. It moves daily. Most finance teams discover how much it moved when the invoice lands.
Before your next renewal, you should be able to answer three questions with evidence rather than impressions:
- Which AI capabilities you are licensed for
- Which are actually switched on
- Who uses them enough to justify the premium
If any answer is a guess, you are funding somebody else’s roadmap.
3. The Cost of Walking Into a Renewal With No Data of Your Own
The first two costs are money leaving the building. This one is more expensive, because it sets how much leaves next year.
Every renewal is a negotiation between two parties. One of them holds precise, timestamped, granular data on how your organization uses the product. The other typically holds a purchase order, a spreadsheet, and a feeling. When you cannot describe your own consumption, the vendor’s account of it becomes the working truth, and every conversation starts from their number.
That number then compounds. Carry 25 percent more entitlement than you need into a contract with a standard annual uplift and you are not overpaying once. You are overpaying slightly more every year, and by year three the overage is simply part of the budget nobody questions.
The same blind spot cuts in the other direction. Audit exposure, true-ups, and compliance penalties all grow from the identical root problem, which is that nobody inside the organization can prove what was actually used. Both risks have the same fix: bring your own evidence.
What Visibility Actually Changes
None of this calls for a transformation program. It calls for measurement that reflects reality.
That is the work Open iT has been doing for more than twenty years. LicenseAnalyzer meters what is genuinely used across more than 6,000 applications, from engineering platforms like Autodesk, Ansys, Dassault Systèmes, and Siemens through to SaaS subscriptions, and turns that into usage trends, chargeback, and evidence you can carry into a negotiation. LicenseOptimizer harvests idle licenses automatically, returning them to the pool before anyone files a request for more.
The outcomes are not theoretical:
- Tata Consultancy Services aligned entitlements with real demand using consolidated global usage data and reported multi-million-dollar savings
- bp separated the Petrel licenses in genuine use from those sitting idle across a complex application estate
- NASA consolidated license servers and strengthened vendor negotiations inside the first year
The Titanic’s ice warnings were never lost. They were received, transcribed, and set aside. The information existed the entire time. It just never reached the person steering.
Your usage data exists too. It is being written right now, in license server logs, admin consoles, and consumption meters nobody has opened this quarter. The next renewal is already on the calendar.
The question was never whether your IT budget is leaking. It is whether anyone reads the warning before the contract renews itself.
よくある質問
What is software shelfware?
Shelfware is software that was purchased but is never opened at all, the licenses sitting completely idle. It is the easiest form of software waste to spot and typically the first thing cleaned up in a cost review. The harder problem is underutilized licenses: seats that do get opened and therefore look active, but see far less real work than their cost implies.
How can you tell if a checked-out license is actually being used?
Checkout data alone cannot answer this. A license shows as “in use” from the moment an application opens until it closes, regardless of how much of that window involved real work. Only activity-level metering, which tracks actual interaction with the application, can distinguish a nine-hour checkout with three hours of real use from a license that is genuinely busy all day.
Why do AI add-ons often go over budget in the first year?
Most organizations buy AI seat licenses for their entire user base upfront, while actual adoption climbs gradually, often reaching only 15 to 35 percent in year one. Combined with consumption-based pricing that bills by token or credit and moves daily, the gap between what is licensed and what is used, and between what is budgeted and what is billed, is where AI spend overruns come from.
How does usage data change a software renewal negotiation?
Without your own consumption data, a renewal is negotiated on the vendor’s terms, since they hold the only detailed record of how the product was actually used. Usage data lets you walk in with your own evidence of what was really consumed, so you can right-size entitlement instead of renewing the prior baseline plus a standard uplift, which compounds the overage every year.
See What Your Usage Data Has Been Trying to Tell You
Open iT meters real software usage across engineering and SaaS environments so you can right-size entitlements, defend renewals with data, and reclaim what you are not using.
