Every company I talk to right now is “doing AI” in some form. New tools rolled out, a few pilots running, maybe a copilot handed to the sales team and a chatbot bolted onto customer support. Good thing, that’s the easy part. Buying the technology was never going to be the hard part.
Here’s the question almost nobody can answer cleanly when I ask it: what has actually changed because of this?
Not what it could change someday. Not what the vendor promised in the demo, with the slide about “40% productivity gains” that nobody has ever managed to reproduce. What’s actually different in business, right now, six or twelve months after you flipped it on.
I get why we’re here. Early on, experimenting made total sense. You don’t know yet where AI genuinely helps and where it’s just expensive noise, so you try a bunch of things and see what sticks. That’s a completely reasonable phase for an organization to be in. But it’s not a phase you get to stay in forever. At some point the experimenting has to turn into results, the same way you’d expect from any other significant investment you put real money behind. I’ve sat in enough budget reviews to know that nobody gets an indefinite pass just because the technology is new and everyone’s excited about it. Eventually someone on the board asks the obvious question, and “we’re exploring a lot of exciting use cases” stops being an acceptable answer.
The mistake I see most often
The mistake I see most often is that teams adopt a tool because it exists and everyone else seems to be getting one, not because it solves a specific, named problem. Flip that order. Start with what’s actually broken, slow, or expensive today. Are responses to customers taking too long? Are people burning hours on work a machine could reasonably do? Is the sales team missing opportunities because nobody has time to follow up? Pick the problem first. Measure where you’re starting from — honestly, with real numbers, not a gut feeling. Then bring in AI specifically to close that gap and go back afterward and check whether it actually did. That last step is the one everyone skips.
And don’t stop the math at the license fee, because that’s rarely where the real cost lives. By the time you add integration work, data cleanup, security review, employee training, change management, and someone who actually owns the thing on an ongoing basis instead of it quietly becoming nobody’s job, the “cheap” pilot usually isn’t cheap at all. It’s worth knowing that full number before you decide to scale something across the whole company.
There’s a risk side to this too, and it’s the one that’s easiest to wave off when things feel like they’re going well. Saving your team a few hours a week doesn’t mean much if it also opens up a data privacy issue, a compliance gap, or a security exposure you now have to manage on top of everything else. I’d rather see a slightly slower rollout with proper guardrails than a fast one that creates a mess for legal or IT to clean up later. Value and risk aren’t two separate conversations — they’re the same one and treating them as unrelated is how organizations end up surprised.
So before your next AI initiative gets a green light, here’s what I’d want straight answers to:
What specific problem is this solving?
Who owns whether it works, by name?
How exactly are we measuring that?
What does it really cost, all in — not just the subscription?
What risk are we taking on, and who’s keeping an eye on it?
If those don’t have solid, specific answers yet, what you’ve got is a project. Not yet a business case. There’s a real difference, and boards are starting to notice it.
We’re past the point where simply saying “we use AI” impresses anyone.
The organizations that actually pull ahead over the next few years won’t be the ones that turned on the most tools the fastest. They’ll be the ones who can sit down, open a spreadsheet, and show exactly what those tools are worth — in dollars, hours, or risk avoided. Everyone else is just going to be guessing, and eventually, guessing gets expensive.
We’re past the point where “we use AI” impresses anyone. The companies that pull ahead won’t be the ones with the most tools switched on — they’ll be the ones who can prove what those tools are worth.
So: if your CFO asked tomorrow, could you actually show the ROI on your AI initiatives — or would you be scrambling?
