Companies have stopped asking how or where they should use AI. That question was settled a long time ago.
Employees across nearly every department are already reaching for AI tools on their own initiative, official rollout or not. The new question they are asking right now is:
“Which of the hundreds of AI use cases for business actually deserves real money?
That’s not a technology question. It’s a capital allocation question.
AI Adoption Is Already Widespread, And That’s a Problem
AI can technically help with almost anything. That sounds great until you realize what it actually means: every team has a wish list, and no company has the budget to fund it all.
Some of those use cases matter a lot. Most don’t. And usage data alone won’t tell you which is which.
Research tracking actual enterprise AI adoption confirms what most leaders already sense: AI use is spreading fast, and it isn’t just new people signing up. The same employees who adopted AI last year use it several times more today than they did then.
So “where can we use AI” is basically answered: everywhere. That was never going to be the hard part.
The hard part is that there are now more plausible AI bets than any company can properly fund. Something has to get cut. The real challenge is deciding which opportunities deserve scarce capital, not finding more places to point AI at.
Why Not All AI Use Cases Deliver the Same Value
Here’s a mistake a lot of companies make: they assume the AI use case with the most users must be the most important one.
It usually isn’t. Something everyone uses a little, like writing help, doesn’t automatically matter more than something a small team uses constantly, like an analyst running deep research every day.
A lot of people doing something small isn’t the same as a few people doing something big.
A company can proudly announce “70% of our people use our AI tool now.” That number says nothing about whether sales went up, whether service got cheaper, or whether the team can handle more work without hiring.
How many people use something and how much value it creates are two different measurements. Mixing them up is how companies end up funding whatever’s popular instead of whatever actually pays off.
Does Higher AI Usage Actually Mean Higher ROI?
It would be convenient if heavier AI usage simply meant a better return. The data even seems to back that up at first glance.
Companies that use AI more intensively tend to post higher revenue per employee and higher value per employee.
But that AI adoption vs ROI cuts the wrong way. The companies using AI most intensively aren’t randomly selected.
The research finds that early, heavy adopters tend to be larger, more valuable, and already more heavily invested in the organizational and intangible capabilities that let new technology stick.
They were probably winning before AI arrived. It’s less “AI made them successful” and more “successful companies use more AI.”
So if usage isn’t a reliable stand-in for value, how to measure the business value of AI becomes the main priority.
To solve it, start with the business, not the tool. Pick a number worth moving:
- Revenue per employee
- Conversion rate
- Customer retention
- Cost to serve
- Customer acquisition cost
- Length of the sales cycle
- Speed of product delivery
- Output capacity per employee
Then ask whether AI can actually move it. The tool comes last, not first.
GFT’s 3 Questions to Ask Before Investing in an AI Use Case
You don’t need a hundred AI ideas.
You need a way to pick the right three or four. AI should compete for investment the same way any other business initiative does: by showing what number it can move, how much that number matters, and whether the impact scales.
Ask this about every candidate:
Question | Why it matters |
What business metric would this actually move? | If there isn’t one, it might still be nice to have. It’s just not worth a big enterprise AI investment. |
How much does that metric matter economically? | A small win on something big beats a big win on something small. |
Does the impact hold up at scale? | Helping one person is nice. Helping hundreds of people the same way is worth real money. |
This isn’t about stopping people from trying things. Try lots of things. Just don’t fund everything that works a little. Fund the few things that actually move a number that matters.
How to Build an AI Investment Portfolio (Not a Wish List)
This is where most companies get it wrong. They add AI use cases to a growing list and hope employees stumble onto the right applications. That’s not a strategy; it’s a wish list with a budget attached.
Once a company has more plausible AI use cases than it can fund well, the mental model has to change. Treat every serious candidate the way a portfolio manager treats a position: sized, tested, and revisited on a schedule, not adopted once and left alone.
If you are wondering how to decide which AI use cases to fund, follow this AI investment strategy:
- Surface what’s already happening. Find where AI is already in use across the organization, including the informal, employee-led usage that almost always predates any official program.
- Score it against the metric it’s supposed to move, and how much that metric actually matters to the business.
- Pilot the strongest candidates at a small scale before committing serious budget.
- Track results against an actual baseline, not against how enthusiastic people feel about the tool.
- Expand what shows real movement. Cut what doesn’t, regardless of how popular it is internally.
The objective isn’t the highest possible adoption number across the workforce. Plenty of companies will post impressive adoption figures and have little financial return to show for it.
The goal is to squeeze the most economic value from the specific bets that keep earning their funding, and to have the discipline to walk away from the ones that don’t.
AI Competitive Advantage: Usage Vs Allocation
It’s tempting to think the company using AI the most will end up winning.
Maybe not. A company where 90% of people use AI for everyday tasks looks impressive on paper. A company where only 30% of people use AI.
But on something that actually drives revenue might be doing a lot better underneath.
Usage numbers don’t tell you who’s winning. They just tell you AI is spreading. Who wins depends on who’s disciplined about where they put it, and that’s a leadership call, not a technology one.
Everyone’s already answered “should we use AI” and “where can we use it.” Here’s the question that actually matters now:
Where can AI move a number big enough to be worth the money?
That question doesn’t care how many tools you have or how many people are logged in. It cares whether something real changed.
The companies that win won’t be the ones using AI the most. They’ll be the ones who got good at picking the right bets. Using more AI was never the advantage. Knowing where it’s actually worth the money is.
Frequently Asked Questions
If a company's AI usage is climbing, does that mean it's getting more value out of it?
Not necessarily. Higher usage often appears alongside stronger financial performance, but that pattern reflects which firms are positioned to adopt aggressively, not proof that more usage causes better results.
Should the AI use case with the most employees using it get funded before a smaller, more intense one?
Not automatically. Reach and intensity measure different things. The better test is whether a use case is tied to a metric that matters to the business and whether it can scale, not how many people happen to be using it.
Why do so many companies report high AI adoption but little financial return?
Because adoption is widespread but scaling is not. Many organizations run pilots or isolated use cases, yet few embed AI deeply enough across core workflows to move enterprise-level earnings.
How long should a use case run before deciding whether to scale it?
There’s no fixed timeline, but it needs to run long enough to compare against a genuine baseline for the metric involved, not just gauge employee enthusiasm. A use case tied to a slower-moving metric, like customer retention, generally needs more runway than one tied to something that shifts week to week, like ticket resolution time.
Is a workflow that only helps one person ever worth enterprise-level funding?
Occasionally, but only if it can be copied elsewhere. A workflow that saves one employee time is a personal efficiency win. The same workflow applied consistently across many employees, accounts, or business units is what turns it into something worth funding at scale.
