Companies of every size buy AI access and treat the purchase itself as adoption. Zylo's 2026 SaaS management index, which tracks tens of millions of enterprise licenses, found that 43% of software licenses go completely unused — an average waste of $80.6 million a year across the organizations analyzed. Flexera's 2025 State of ITAM Report found that 35% of companies saw that waste grow over the past year, even with more teams watching it.
When the product bought is AI, the same pattern repeats and gets more expensive to ignore. IBM's Institute for Business Value study, "Rewiring the C-Suite," surveyed more than 2,000 chief executives and found that only 25% of employees use AI regularly at work — inside companies that already bought and rolled the tool out to the whole team. The question that matters to whoever signs the budget isn't how many AI licenses the company has. It's how many of them anyone actually opens.
That matters because the cost leaves the account the same month the license is signed, and the return, when it shows up, gets lost in a usage average almost nobody checks person by person. An idle seat costs the same every month, whether or not it produces anything.
The Cost of an Unopened License
Research from Productiv, which tracked nearly 100 million software licenses over three years, landed in the same range as Zylo: 40% of enterprise software licenses go unused. This isn't a failure of one vendor or one category of tool — it's a structural pattern in how companies buy technology: first by forecasted need, then by renewal habit, almost never by confirmed real use.
With AI, the same pattern shows up with a costlier twist: the gap between having access and actually using it is wider than in other software categories. BCG's 2025 research on AI at work called this gap a glass ceiling: more than three-quarters of leaders and managers use generative AI several times a week, while regular use among frontline employees has stalled at 51%. The tool reaches everyone; the habit of opening it doesn't.
The IBM study cited above spells out the other side of that same gap: even with broad access already rolled out company-wide, only 25% of employees use AI regularly at work. Between an active account and a formed habit, most companies still have nothing that carries people across.
The reason is rarely lack of interest. It's the absence of three concrete things: a routine where AI is already expected, a ready example of how that specific task gets solved with it, and certainty that doing so is allowed. Without those three pieces, opening the tool turns into a guessing game — and most people would rather finish the task the way they already know how.
Access Is Not a Routine
The most common response to low usage is handing out more seats or scheduling a one-off training session. Neither does much alone. BCG measured this directly: only about a third of employees receive real training, and those who get at least five hours of training with hands-on coaching show clearly higher regular use than those who only got a license.
Slack's Workforce Index, published in late 2024, showed the other side of the problem: US adoption practically stopped growing in one quarter — from 32% to 33% — and nearly half of desk workers (48%) said they felt uncomfortable admitting to their own manager that they'd used AI for a common task. Christina Janzer, who leads Slack's research team, summed up the problem in one line.
A lot of the burden has been put on workers to figure out how to use AI on their own.
That burden is what explains the unopened license. It isn't that the person decided not to use AI — it's that nobody decided for her when she can use it, for what, and how far she can go. Without that decision made in advance, every attempt starts from zero and carries the risk of getting it wrong alone. Most people choose not to take that risk.
What Has to Be in Place
An AI environment a team actually uses is defined by verifiable mechanisms, not one more adoption campaign.
Corporate identity, not a standalone account. Access comes from the same directory that already controls email and internal systems, with no parallel account for someone to remember to deactivate when a person changes roles or leaves.
Access matching each person's role. Each person sees only the tools and actions their own role authorizes — not everything, not nothing, just enough for their own routine.
Human approval matched to risk. Sensitive actions pause and ask someone to confirm before going ahead. Documents that will become official knowledge go through a two-step review and approval, with the person who wrote it separate from the person who approves it.
An audit trail. Every relevant use gets logged — who did what, when, and with what result — so an audit question gets answered in minutes, not weeks.
Reuse with permissions. When someone on the team builds an agent, a conversation script, or a sequence of steps that solves a task well, it can be made available to other people, groups and roles, without turning into unrestricted access for people who shouldn't have it.
That last point is what closes the loop between access and use: without it, the routine that worked stays locked inside whoever found it, and the next person starts from zero again. That's how Skyller was designed: identity from the company's own directory, permission by role, and reuse of what already works, as the default — not as a side setting.
From Individual Licenses to Shared Use
Under the per-person license model, the AI budget behaves like any other poorly sized software line: part of the team never comes close to its own account limit, while another part hits it every week. Neither shows up in the finance report — only the total spend, with no indication of whether it turned into work.
When access is shared across the team instead of locked per person, the same budget stretches further: whoever needs more in a given week uses more, whoever needs less frees up room for the team, and consumption becomes visible by department — not one single "AI subscriptions" line lost among a hundred others.
The bigger gain, though, is what happens to the work itself. A routine discovered by one person stops dying with her. It becomes something the next person reuses, adjusts and improves — and the company starts accumulating capability, not just spend.
Before You Renew the Next License
Before signing another year of AI access, it's worth bringing these questions to the budget meeting with IT and operations leadership:
- How many AI accounts does the company pay for today, and how many were opened in the last week? If nobody can answer, the problem isn't the vendor — it's the lack of visibility into the spend itself.
- When someone on the team finds a good way to use AI on a task, does it become something another person can reuse? If the answer is "only she knows how to do it that way," the company is paying for the same discovery again, quietly, with every new hire.
- Does a new employee know, on day one, what they can and can't ask AI to do? If the answer depends on asking a coworker, the license was never given a routine.
- Does AI cost show up broken out by department in the budget, or as one lump line? Without that breakdown, nobody can tell whether one team barely uses it while another overuses it.
