Pull up your AI spend for last quarter. You can probably find it in about thirty seconds — a line item, a per-seat number, a total you feel reasonably good about.
Now find the rest of it.
The subscriptions on personal cards, coming back through expense reports as “software,” or “professional development,” or nothing at all. The department that signed up for its own tool because waiting for IT wasn’t an option. Four people on your team paying twenty bucks a month each for four different AI models, all doing roughly the same thing, none of which your company actually owns.
That’s the second bill. It doesn’t arrive. Nobody approves it. And it is not the expensive part.
The expensive part is what doesn’t compound
Here’s the thing that should bother a Salt Lake City finance leader more than the duplicate subscriptions.
In PwC’s 2026 Global CEO Survey, 56% of chief executives said AI had produced no significant financial benefit for their business. Only 12% could point to both a cost savings and a revenue improvement. These aren’t companies that failed to adopt AI. These are companies that adopted it, spent real money on it, and still can’t find where the return went.
Here’s why. Four people using four different AI tools aren’t four times as productive. They’re four silos. Someone in operations figures out how to cut a recurring three-hour task down to forty minutes — and that knowledge stays in their head, in their tool, on their personal account. Someone in finance solves an almost identical problem six weeks later, starting from zero, in a completely different product. Nothing accumulates. Nothing transfers. Nothing gets better across the company just because it got better for one person.
You’re paying for individual productivity and getting exactly that: individual, unrepeatable, unmeasurable productivity that walks out the door the day that person leaves.
That’s before we even get to the parts that don’t show up as spend at all. Company data sitting in tools you don’t control. No volume leverage on pricing, because you’re buying in ones instead of in blocks. Per-seat rates a fraction of what you’d negotiate as an organization. And no line of sight — none — from any of it back to a recovered hour or a recovered dollar.
This is what makes AI sprawl uniquely bad as a category of spend. Most uncontrolled spend is just wasteful. This kind is wasteful and it’s compounding against you, because the longer it runs, the more the value gets locked into places you can’t reach — and the more exposure builds underneath it. We’ve seen the real cost of that exposure firsthand: organizations running with high shadow AI levels see an average of $670,000 added to the cost of a data breach when one happens, pushing the average breach north of $4.6 million. That’s not a hypothetical. That’s the bill sitting behind the bill you can already see.
The five numbers you can’t produce
Try this. Right now, without asking anyone, answer these five:
- How many AI tools are actually running in your business? Not licensed. Running. Including the ones on personal accounts.
- What percentage of your workforce uses AI weekly — not “has tried it once,” but uses it every week?
- What percentage of that usage is governed — a sanctioned tool, under your control, with a policy behind it?
- What percentage is charged back to a department, a cost center, a budget owner?
- What percentage is tied to a stated business outcome, rather than one employee’s personal preference?
If you’re like most SMBs we work with in Salt Lake City, you can’t answer any of the five with confidence. Some leaders can’t answer the first one within a factor of two.
Sit with that for a second. This is a category of spend and risk running inside your business right now — touching client data, absorbing budget, shaping how work actually gets done — and there’s no number attached to any of it. You wouldn’t accept that in any other line of the business. You’d have found it in a review a long time ago.
The reason it hasn’t been found is that nobody owns it. It didn’t come in through procurement. It came in through people trying to do their jobs.
You can’t fix what you can’t see
Notice that none of this is an argument for spending more. It’s an argument for seeing what you’re already spending.
That’s the good news buried in here, and it’s the reason this is worth an hour of your time rather than a project plan. Most companies that finally look at these five numbers discover two things at once: they’re spending more than they thought, and they’re capturing less of it than they thought. Which means the first move isn’t a purchase. It’s consolidation — one sanctioned path, volume pricing, gains that stack instead of scatter, and a policy that gives your people permission to use the thing you’ve actually already paid for.
We’ve watched this play out with clients across Utah for over 30 years, long before “shadow AI” was a phrase anyone used: companies routinely find they were already paying for a capability inside tools they hold today, while separately reimbursing staff for personal subscriptions to do the same work. That’s not a technology problem. That’s a visibility problem — and visibility is cheap.
The hours are there. Most SMBs are sitting on four to eight hours per knowledge worker, per week, in repetitive admin and information search. The spend is there too. Right now, neither one is on a report you’ve seen.
Come find your five numbers
We’re running a working session on exactly this: what shadow AI is costing SMBs, how to find the five numbers inside your own business, and what to actually do with them once you have them.
Not a product demo. Not a keynote about transformation. A method for producing the numbers, and a walkthrough of what most companies find when they run it.
Webinar: The Shadow AI Audit — What’s Really Running in Your Business
45 minutes. You’ll leave with the five-number framework, the questions to ask your own team, and a clear read on the cost and exposure sitting inside your company right now.
Built for finance and operations leaders at companies between 25 and 200 people.
Can’t make it live? Register anyway and we’ll send you the recording and the audit worksheet.
Questions before then? Call us at (801) 562-8778 or email [email protected] — we’ve been untangling IT problems for Utah businesses since 1992, and this is just the newest version of an old one.
