Across industries and across the globe, there’s no shortage of investment going into AI technology and training right now. But we’re seeing more organizations struggling to determine what they’re actually getting back from it.
You can see some of that struggle in the numbers. Already, 65% of companies have abandoned AI initiatives because of gaps in skills and execution. By the time an initiative gets shelved, an organization may have invested in the technology, implementation, and training, along with months of employee time spent evaluating it, testing it, and figuring out how it fits into the work. That investment doesn’t disappear from the balance sheet, but the return (oftentimes never clearly defined to begin with) it was supposed to generate might.
And not every cost is tied to an initiative that gets abandoned. Some are much harder to spot on a budget: the hours employees spend correcting AI-generated work, the tools that were purchased but barely get used, or the processes that got an AI upgrade without becoming noticeably better.
Individually, those things may not look particularly expensive. Spread them across teams, workflows, and an entire organization, though, and they start telling a very different story about what your AI investment is actually costing you and your people.
AI can save time. It can also create work.
The promise of AI efficiency usually starts at the task level. Something that took an hour now takes 15 minutes. A blank page becomes a first draft, or a mountain of information gets condensed into a summary. On paper, the time savings can look pretty compelling. But a faster task doesn’t always mean a faster workflow.
A first draft generated in minutes may still need significant editing before it’s usable. An employee might spend 20 minutes prompting their way to an answer they could have reached themselves in 10, while someone else completes their part of a process faster with AI only to hand off work that requires more review downstream.
None of that means AI failed—it just means you have to follow the work a little further before deciding how much time or effort it actually saved. Look beyond the individual task and ask:
- Did the work move faster from beginning to end?
- Did the quality improve along with the speed?
- Was someone’s expertise freed up for something more valuable?
- Did AI actually eliminate work, or simply move it somewhere else in the process?
Sometimes AI will make a task dramatically easier. Other times, using it creates more work than it removes, and sometimes the smartest use of AI is no AI at all. The real return on your AI investment comes from you and your workforce knowing the difference.
More AI doesn’t always mean more efficiency
That calculation gets a lot more complicated when you move beyond one task or employee and look across an entire organization.
One team finds a tool it loves while another buys something with similar capabilities. Employees develop their own prompts, processes, and workarounds, and different departments spend time solving versions of the same problem without realizing another team may already have a perfectly good (or even better) solution.
Some of that is a natural and necessary part of figuring out where AI belongs. People need room to experiment, learn what works in the context of their roles, and discover where the technology genuinely improves the work.
What matters is what happens to those experiments afterward. Without a clear way to share what works, learn from what doesn’t, and turn useful practices into repeatable ones, organizations can end up paying for overlapping tools while employees repeatedly rebuild knowledge that already exists somewhere else in the business.
At that point, scaling AI doesn’t necessarily mean scaling its value. The same systems that can spread a great use case across hundreds of employees can just as easily spread an inefficient one.
The workflow might be the problem
There’s a particularly telling gap in how organizations are adopting AI: 75% of teams use it regularly, but only 38% have redesigned workflows around it. That leaves a lot of room for AI to become part of the work without fundamentally changing how the work gets done.
Imagine a process with six steps where AI suddenly makes step three twice as fast. There’s an obvious efficiency win there, but looking at the process as a whole might reveal a much bigger one. Maybe steps two and three could now be combined, step four no longer needs to happen at all, or there’s a better way to start the process entirely. In other cases, adding AI somewhere in the workflow may create enough additional review that it isn’t saving much time at all.
Those opportunities are easy to miss when the focus stays on finding individual tasks where AI can help, rather than looking at how the work itself could change.
That means asking bigger questions about the workflow:
- What can disappear entirely?
- Where can AI genuinely accelerate the work?
- Which decisions still depend on human context and judgment?
- Which handoffs no longer make sense?
- Where might AI be making something more complicated without making it meaningfully better?
There’s little value in adding AI to a workflow simply because you can. Getting more from the investment means looking beyond how many workflows use it and asking whether those workflows are actually better for it.
Small inefficiencies don’t stay small
Five extra minutes correcting an AI output probably won’t set off any alarms. Neither will an unused license, a redundant tool, or a pilot that quietly fades away after a few months. But spread those inefficiencies across hundreds or thousands of employees, and the math starts to look different.
Much of the waste won’t show up anywhere as an obvious AI expense. It shows up in employee time spent reworking an output, managers reviewing something that wasn’t quite ready, technology spend scattered across teams, or processes that take longer than they should. Even an AI initiative that never makes it into day-to-day operations can consume months of attention and resources before anyone decides to move on.
None of that will show up as a neat line item on an AI budget, but it still affects the return.
Which is also why adoption numbers can only tell you so much. Knowing that employees are using AI is useful, but knowing whether that use is producing better work, stronger workflows, and measurable business outcomes gives you a much clearer picture of whether the investment is actually paying off.
Protect the investment you’ve already made
It’s easy to question the value of spending more on AI training when the investment in the technology itself has yet to pay off. But getting more value from that investment depends, at least in part, on whether people know how to use AI well, which goes beyond simply knowing how to write a better prompt.
Employees need the skills to recognize where AI can genuinely improve their work, evaluate what it produces, and know when their own expertise should take the lead. That means building practical AI capabilities across teams that are grounded in the work your people actually do, rather than treating AI as a standalone skill or technology.
Leaders have a different role to play. They need enough hands-on AI experience to make informed decisions about where AI belongs, where it doesn’t, and what needs to change around it, including when the workflow itself needs to be rethought.
Over time, those capabilities give organizations something they can actually build on. Useful practices can spread instead of being rediscovered team by team, lessons from failed experiments can inform the next one, and decisions about additional AI investment can be based on a much clearer understanding of where it creates value.
Get more from what you’re already spending
The return on AI won’t be determined by your technology budget alone. It’ll be shaped by hundreds of decisions about how that technology actually gets used: which workflows change, which don’t, what people know how to do with it, and whether useful experiments ever become better ways of working.
So before adding another tool, license, or platform to the budget, it may be worth taking a closer look at the investment you’ve already made and asking a much simpler question:
How much value are you actually getting from the AI you already have?
And if you need help figuring out the answer, we’re here and ready to help.