Why most businesses should diagnose friction before buying AI

Most businesses buy AI the way they buy insurance: because everyone else seems to have it, and going without feels risky. That instinct is understandable, but backwards.

Tools solve problems. If you don't know precisely what the problem is, a tool is just an expensive guess. The better sequence is to find the friction first, then decide what fixes it, if anything.

What “diagnosing friction” means

Friction is anything that costs time, money, or attention without adding value. It shows up as a task someone dreads, a spreadsheet that gets rebuilt every Monday, or a customer question that takes three people to answer. It is rarely labeled. Nobody puts “manual data re-entry“ on a whiteboard as a strategic priority. It happens every day, in the background.

A proper diagnosis means watching how work flows, not how the org chart says it should. It means asking the person doing the task, not just the person managing them. And it means writing down what you find in hours and dollars, not vague impressions.

Two ways this goes wrong

Picture a small accounting firm that buys a chatbot to “handle client questions“ after a partner reads about it in a newsletter. The real issue turns out to be that client documents arrive in six different formats and someone has to manually sort them before any work can start. The chatbot never touches that problem. The firm is now paying for two things: the original friction, and a subscription.

A growing e-commerce brand hires a vendor to build a custom agent for order support. Only after the contract is signed does the team learn that most support tickets trace back to a confusing return policy on the website. No agent fixes a policy. A paragraph of clearer copy might have.

In both cases, the company skipped a step. They matched a tool to a feeling of inefficiency instead of a diagnosed cause. The fix always arrives late, and it's often the wrong fix entirely.

Questions to ask before you buy anything

• Can I describe the friction in one sentence, with a number attached: hours per week, dollars per month, tickets per day?
• Have I watched the actual task being done, or am I relying on a secondhand description of it?
• Is this a technology problem, a process problem, a training problem, or a decision that keeps getting deferred?
• If I fixed the process by hand, with no new software, would most of the pain disappear anyway?
• Am I buying this because a competitor has something similar, or because I have evidence it addresses my specific bottleneck?

None of these questions require technical expertise. They require patience and a willingness to look closely at unglamorous, everyday work before reaching for a purchase order.

Why this order matters commercially

Diagnosing first is a financial decision, not a matter of being careful. A tool bought to solve the wrong problem still costs money every month, still needs someone to maintain it, and still leaves the original friction untouched. That is a worse outcome than doing nothing, because now you're paying twice: once in lost time, once in software fees.

The businesses that get value from new technology are the ones that can point to the exact task it replaced and the exact hours it gave back. That clarity only comes from looking first and buying second.

AI is a powerful answer to a narrow set of well-defined problems, and a poor one for a vague feeling that things could run better. Before evaluating any tool, find out where friction lives in your business: in which task, on which day, at what cost. The diagnosis usually tells you what to buy, if anything, and skipping it only moves the guesswork from before the purchase to after.