The Tokens Were Never the Problem
What Should a €25m to €250m Company Spend on IT? — the benchmark data pulled together across fifteen industry verticals, with sources shown. Find your sector, find your number.
Download the benchmark report (PDF)
Token costs are the loudest conversation on LinkedIn right now. They're also the wrong one.
If you've arrived here from that conversation, here's the longer version of the argument, with the numbers underneath it.
The pattern
A company reaches out wanting to "do AI". Sensible ambition, no argument from me. But before anything else, we look at one number: current IT spend as a percentage of annual revenue.
The mid-market median sits around 4 to 8% depending on sector. The company in front of me is at 0.8%, in a sector where 2 to 4% is the floor. They're not an outlier. They're the pattern.
The plan, usually unspoken, is that AI will be the shortcut. Skip the years of unglamorous investment in connected systems and clean data, and jump straight to the clever part.
I understand the appeal. Nobody got excited budgeting for systems integration. But the data tells a clear story about why the shortcut doesn't exist.
What the benchmarks actually say
When you pull together the major benchmark sources, Gartner's midsize enterprise data, Flexera's industry surveys, Deloitte's mid-market research, two findings stand out for any leadership team in the €25m to €250m range.
First, the spread between sectors is enormous. Gartner's data for companies under $250m in revenue puts the cross-industry average at 6.2% of revenue. But underneath that average, food and beverage processors sit around 1%, industrial manufacturers around 2 to 4%, and software businesses at 10% and beyond. Flexera's survey puts software companies as high as 18%. Knowing the all-industry average tells you almost nothing. Knowing your sector's range tells you nearly everything.
Second, and this is the one that surprises boards: mid-market companies should be spending a higher percentage than the giants, not a lower one. Large enterprises with five thousand or more employees average around 3.7% of revenue on IT. Mid-market firms typically need 6 to 8%. That's not inefficiency. A €50m company needs much of the same core machinery as a €5bn one, ERP, security, integration, but spreads the cost over far less revenue. So when a mid-market leadership team benchmarks itself against enterprise headlines and concludes it's "running lean", it's often measuring with the wrong ruler. Lean and under-built look identical on a budget line.
The amplifier problem
I play guitar. Badly enough that "Wish You Were Here" at parties is the ceiling. But I learned early that a better amplifier doesn't fix a badly tuned instrument. It makes the tuning problem louder, and more expensive to listen to.
AI is an amplifier. Point it at a well-connected operation and it compounds everything good about it. Point it at fractured systems, half-migrated data and spreadsheet workarounds, and you get the same fractures, faster and at volume.
This is why so many mid-sized firms are now staring at serious token bills for AI projects with no ROI to show. The model isn't failing them. It's faithfully amplifying an operation that was never wired to be amplified. The tokens were never the problem. They're the invoice for the foundations that were skipped.
What catching up actually costs
Here's the part the benchmark report covers that rarely makes it into the AI conversation: the difference between run-rate spend and transformation spend.
The 4 to 8% ranges are steady-state numbers, what a digitally healthy company spends to stay healthy. A company catching up from years of under-investment should expect to run above its sector range, often in the 6 to 10% territory, for a two-to-five-year stretch while the foundations go in. That's normal. It's in the data. Boards that treat it as a temporary, deliberate overshoot fund it properly. Boards that don't are the ones who reach for AI as the cheaper alternative, and end up paying for both.
So the more useful question isn't "what's our AI strategy?" It's "are we even keeping up with our competition on the basics?" If the answer is yes, AI will likely compound that advantage faster than anything you've invested in before. If the answer is no, that's not a verdict. It's just the real starting line.
Check your number
Do you know your IT spend as a percentage of revenue, and where your sector's range sits? No worries if not. We've pulled the benchmark data together for companies with annual turnover between €25m and €250m, broken down across fifteen industry verticals, with the sources shown. Free download, no email gate: download the benchmark report.
Find your sector, find your number, and see which side of the range you're on. If you're below the floor, that's the conversation to have before the AI one.
And if it turns out your AI budget is currently going into someone else's guitar case, you know where to find us.