At my former agency, we tried something like a dozen content management systems over the years. Some of that was honest curiosity. We liked experimenting. But a lot of it was a search for the one that would finally be easy for clients to use and wouldn’t take us forever to set up. Some really did handle content better than others. None of them fixed the real problem, because the problem wasn’t the software. It was content strategy: deciding what content a client actually had, how it should be organized, and who would keep it current. Without that, it was all for naught. Every new system just gave the same mess a nicer admin screen.
You’ve probably lived your own version. Jobs keep slipping, so the team spends a month picking a new project management tool (“I’ve heard good things!”) and another month moving everything over. Ahh, this feels different. Three months later, jobs are still slipping, now with prettier notifications. Now with AI! The real problem was that nobody had decided who covers a job when its lead is out, and no new tool was going to make that call.
It’s the balance-transfer move. The debt doesn’t shrink. It just moves to a card with a better introductory rate. Switching tools works the same way: the mess moves to a new home, and the decision you were avoiding comes along for the ride.
A tool runs whatever process you give it. If the real problem is a decision nobody has made (who owns a step, what “done” means, when a client’s feedback is final), software can’t make it for you. It just carries out the mess faster and charges you per user to do it.
Diagnose first. Then buy.
What “diagnose before you buy” means
Diagnosing can be as simple as naming the problem in one sentence and finding the step where the work actually breaks. “We need better project management software” is a Band-Aid. “Design reviews wait four days because nobody owns the client approval” is a diagnosis.
The difference matters because symptoms and causes call for different fixes. Late jobs, missed handoffs, a shared drive nobody can search, invoices that go out two weeks after the work is done: those are symptoms. Underneath each one there’s usually a missing owner, a missing rule or a missing definition. A tool can hold the rule once you’ve written it down. It can’t write it.
What the research says
This is an old problem. In 1990, Michael Hammer wrote in Harvard Business Review that companies “tend to use technology to mechanize old ways of doing business,” and told them it was time to “stop paving the cow paths.” His best-known example was Ford. Its North American accounts payable department employed more than 500 people, and Ford hoped computers would get that down to about 400. Then it looked at Mazda, which ran accounts payable with five. Once Ford rebuilt the process itself instead of speeding up the old one, Hammer reported a 75 percent cut in head count.
Software buyers keep relearning it. In Capterra’s 2024 Tech Trends survey of 700 software buyers at United States businesses, 58 percent said they regretted at least one software purchase from the previous 12 to 18 months. Asked what they’d change next time, the two most common answers were about the decision, not the product: getting everyone to agree on how to judge the options (38 percent) and clarifying goals and outcomes (37 percent). That’s another way of saying they wished they’d named the problem first.
AI tools are following the same pattern, only faster. A 2025 report from a research project at the Massachusetts Institute of Technology’s Media Lab found that 95 percent of organizations were getting no business return from their generative artificial intelligence spending, and it pointed to brittle workflows and “misalignment with day-to-day operations” among the reasons, according to Virtualization Review’s coverage. That study looked at large companies, not 20-person firms. The lesson carries over anyway: drop a tool onto a process nobody has defined and you get very little back.
These are a case study and two surveys, not controlled experiments, so treat the numbers as direction, not precision. The direction is consistent.
Signs you’re buying a tool to dodge a decision
Nobody can say in one sentence what will be different six weeks after launch.
The problem is really about who decides something, like who approves estimates or who can say yes to a client’s extra request, and the tool is supposed to settle it without anyone having to say it out loud.
You already bought a tool for this same problem once.
The pitch inside the company is some version of “once we have this, everyone will finally…”
And the setup screens ask questions you can’t answer.
That last sign is the most useful one. Every project tool asks you to define your stages. Every client database asks what counts as a qualified lead. Every time-tracking tool asks which tasks are billable. If your team argues about those answers, the argument is the decision you’ve been putting off. The tool just put it in front of you.
A test to run before any purchase
Before you sign up for anything, answer five questions. If you can’t, you’re not ready to buy.
- Can you write the problem in one sentence without naming a product?
- Can you point to the last three times it happened, and the step where each one broke? If all three broke in different places, you may have three problems. If the same person was in the middle of all three, read whether it’s a people problem or a process problem before you buy anything.
- Who owns that step? If the answer is “nobody” or “everybody,” naming an owner is the fix, and it’s free.
- Could you run the fixed process on a spreadsheet or a whiteboard for two weeks? If it works there, buy a tool to make it faster or to handle more volume. If it doesn’t work on paper, software won’t rescue it.
- What number should move, and when will you check? Hours spent on a task, days from approval to invoice, jobs delivered late. Write the number down now, before the tool arrives, so you can tell later whether it earned its subscription.
When buying the tool is the right call
Sometimes the tool is the answer. If the process works and you’ve simply outgrown the spreadsheet, buy the tool. Same if a client contract or a security requirement calls for a specific kind of system, or if you’ve already made the decision and want software that enforces it, like a required field that stops a job from closing until the final invoice is out.
A cheap, short trial can be part of the diagnosis too, as long as you treat it that way: give it an end date and one question to answer. On a team of three or four people, the owner usually is the process, so the diagnosis might be a single conversation. Have it, then buy.
My own rule of thumb is one tool in, one tool out. It doesn’t have to be a straight swap. A new project tool doesn’t mean cancelling some other project tool. But every time something new comes in, something has to go. Software subscriptions pile up the same way streaming services do, one reasonable monthly charge at a time, until you open your profit and loss statement and wonder what half of them are. The rule isn’t really about the count. It forces a look at everything you’re already paying for before you add one more.
Where to start
If you can’t pass the five-question test above, that’s your starting point, not a reason to keep shopping.
That’s the job of The Operations Diagnosis: a month inside the business mapping how work flows and ranking the three problems costing you the most. Sometimes one of them needs a tool. More often, the top one needs a decision.
Name the problem. Then go shopping.
