A data analysis of how much software companies now run outside the IT department, who actually holds the buying power, what the unmanaged sprawl costs, and why “shadow AI” turned a governance headache into a breach-cost line item.
Shadow IT makes up 45% of the average organization’s applications in 2026, according to Zylo’s SaaS management data — nearly half of every company’s software runs without IT owning it. That is not an edge case or a rogue-employee problem; it is how modern software gets bought. The numbers below show how big the shadow portfolio has grown, who controls it, what it wastes, and why the new front — unsanctioned AI — now shows up directly in breach costs.
Key takeaways
- Shadow IT is 45% of the average company’s applications in 2026, per Zylo — up from a category that has hovered near half of the portfolio for years
- Business units control 81% of SaaS spend; IT directly manages just 15%, per Zylo’s 2026 SaaS Management Index
- Small businesses run 68% of their apps as shadow IT, versus 52% at midsize and large companies, per Productiv
- 78% of AI users bring their own AI tools to work — 80% at small and midsize firms — per Microsoft’s Work Trend Index
- One in five breached organizations (20%) was compromised through shadow AI, adding $670,000 to the average breach, per IBM’s 2025 Cost of a Data Breach report
- 63% of breached organizations have no AI governance policy or are still writing one, and 97% of those breached through AI lacked proper access controls (IBM)
- 67% of expensed applications carry a “Low” or “Poor” risk score, and 51% of software purchases are miscategorized in expense reports — the seams where shadow IT hides (Zylo)
How much of a company’s software is shadow IT?
Roughly half. Zylo puts shadow IT at 45% of the average organization’s applications in 2026, and Productiv found unmanaged apps made up 56% of the portfolio, with the average company carrying about 142 shadow IT apps. Whichever figure you take, unsanctioned software is not the exception — it is close to the median.
The split widens by company size. Productiv reports small businesses run 68% of their apps as shadow IT, against 52% at midsize and large firms, because smaller teams rarely have the procurement gatekeeping that slows down a departmental credit-card purchase.
Who actually buys software now — IT or the business?
The buying power has moved. Zylo’s 2026 index found lines of business control 81% of SaaS spend while IT directly manages just 15%. In large enterprises, expensed applications — the ones charged to a corporate card and reimbursed — account for 41% of the software stack, though less than 1% of total spend.
That inversion is the mechanism behind shadow IT. When a marketing lead or a sales team can activate a tool in minutes and expense it, the approved software catalog stops describing what the company actually runs. IT inherits the risk without ever seeing the purchase.
What does shadow IT cost?
The waste is structural, not incidental. Zylo found organizations leave 36% of their SaaS licenses unused, with median SaaS spend of $9,455 per employee. When nearly half the portfolio is bought outside IT, no one reconciles renewals, deduplicates overlapping tools, or reclaims idle seats.
Redundancy compounds the bill. Expense-based SaaS spending rose 267% year over year in Zylo’s data, and ChatGPT is now the single most-expensed application — a signal that the fastest-growing shadow spend is landing on individual cards, one subscription at a time, below the threshold that triggers a procurement review. The cost picture mirrors what we found in our SaaS spend statistics.
How risky is unmanaged software?
Very, and the risk is invisible by design. Zylo reports 67% of expensed applications receive a “Low” or “Poor” risk score, and 51% of software purchases are miscategorized in expense reports — the exact blind spot where shadow IT accumulates. Productiv adds that shadow apps average just 2.3 compliance certificates against 3.9 for IT-owned tools.
Every one of those unvetted apps is a credential store, an OAuth grant, and a data-processing agreement no one reviewed. The security case for consolidation is the same one behind dedicated password managers and the identity hygiene we cover in our password security statistics: fewer accounts mean less to breach.
What is shadow AI, and how common is it?
It is the fastest-moving branch of shadow IT. Microsoft’s Work Trend Index found 75% of knowledge workers now use generative AI at work, and 78% of them bring their own AI tools — rising to 80% at small and midsize companies. Employees are pasting source code, contracts, and customer records into tools IT never approved.
The behavior is deliberately quiet. Microsoft found 52% of workers are reluctant to admit using AI for their most important tasks, which is precisely why shadow AI evades detection: the people using it have a reason not to report it.
How much does shadow AI add to a data breach?
A measurable premium. IBM’s 2025 Cost of a Data Breach report found 20% of breached organizations were compromised through shadow AI, and those with high shadow-AI exposure paid $670,000 more than peers with little or none — against a global average breach cost of $4.44 million.
Shadow-AI breaches also hit more sensitive data. IBM found they compromised customer PII in 65% of cases and intellectual property in 40%, versus global averages of 53% and 33%. The unsanctioned tool is not just an extra door; it routes the crown jewels through it.
Why can’t IT see any of this?
Because the controls were never built. IBM found 63% of breached organizations had no AI governance policy or were still developing one, and among those breached through AI, 97% lacked proper access controls. Only 37% have any policy to manage AI or detect shadow AI at all.
Governance also lags usage. IBM found just 34% of organizations with an AI policy actually audit for unsanctioned AI. A policy that no one enforces is indistinguishable, to an attacker, from no policy — which is how a tool nobody approved becomes the breach nobody saw coming.
Frequently asked questions
What percentage of company apps are shadow IT?
About 45% of the average organization’s applications, per Zylo’s 2026 data, while Productiv measured 56% and roughly 142 shadow apps per company. Small businesses run the highest share at 68%. The category has hovered near half the portfolio for years.
Who controls software buying today?
Lines of business, not IT. Zylo’s 2026 SaaS Management Index found business units control 81% of SaaS spend while IT directly manages just 15%. In large enterprises, expensed applications alone make up 41% of the software stack, which is why so much software is invisible to central IT.
How much does shadow IT waste?
Organizations leave 36% of their SaaS licenses unused at a median $9,455 in SaaS spend per employee, per Zylo. Because shadow purchases skip renewal reviews and deduplication, redundant and idle subscriptions accumulate unchecked — expense-based SaaS spending alone rose 267% year over year.
What is shadow AI?
Shadow AI is the unsanctioned generative AI tools employees adopt without security sign-off. Microsoft found 78% of AI users bring their own tools to work. Those tools ingest code, contracts, and customer data, then send it to third-party providers outside corporate control.
How much does shadow AI increase breach costs?
By about $670,000. IBM’s 2025 report found 20% of breached organizations were compromised through shadow AI, and high-exposure firms paid $670K above the global average breach cost of $4.44 million. These incidents compromised customer PII in 65% of cases.
Why is shadow IT so hard to detect?
Because it hides in the accounting. Zylo found 51% of software purchases are miscategorized in expense reports and 67% of expensed apps carry a low or poor risk score. On the AI side, 63% of breached firms had no governance policy at all, per IBM.
What this means for teams
Shadow IT is the strongest possible argument for the 80/20 approach. When nearly half your apps are unmanaged, business units hold 81% of the spend, and two-thirds of expensed tools score poorly on risk, the problem is not that employees are careless — it is that the portfolio is too large to govern. Every unvetted tool is a login, a data pipe, and a renewal no one owns.
The fix is not a longer approval queue; it is a shorter list. A consolidated stack of fewer, better-chosen tools shrinks the attack surface, kills the redundant spend, and gives IT something small enough to actually secure — including a sanctioned answer to the AI tools employees will otherwise bring on their own. That is the whole thesis behind how we score every pick on the about page: fewer, better tools beat a sprawling shadow portfolio every time.