Here’s a diagnostic question I ask early in AI adoption conversations: think about the last significant technology initiative your business ran — not AI, anything. Can you tell me what the measured outcome was?
If the answer is “yes, we set a baseline of X, we targeted Y, we achieved Z” — congratulations, your business has the muscle to measure AI outcomes too. That muscle is rare.
If the answer is “it went well, I think” or “we’re still figuring out the ROI” — that’s the same answer you’ll give about AI in two years. The measurement problem isn’t specific to AI. It’s a general capability your business either has or doesn’t. And AI adoption will not build it for you — if anything, AI initiatives make the measurement problem worse, because the technology moves faster than the measurement discipline.
This isn’t a criticism. Most SMBs don’t rigorously measure the outcome of technology initiatives, because the incentive structure doesn’t reward it. Nobody’s bonus depends on producing a clean before/after comparison. Vendors don’t want you to have baselines because baselines let you see how much of their promise was marketing. Executives who championed the initiative have motivated reasoning against too much honest measurement.
The fix isn’t complicated: for the next initiative — whether it’s AI or not — write down the baseline and the target before you start. Attach specific numbers. Assign someone to measure the outcome six months later. Publish the result inside the business, whether it’s flattering or not.
Do that once, honestly, and you’ve built a measurement muscle that will carry over to every future initiative. Skip it, and you’ll be having the same “what worked?” conversation in three years about AI that you’re probably having now about your last CRM implementation.
Here’s the deeper point: the measurement discipline isn’t about AI. It’s about being a serious organization that can tell whether its decisions are working. AI is the accelerant that reveals this capability or its absence. Every SMB that adopts AI without a measurement discipline ends up in one of two places — a growing collection of unmeasured tools that no one can defend, or a growing suspicion that AI doesn’t work for them. Neither is true. Both are the symptom.
The diagnostic asks: for your last major initiative, can you produce the numbers? Answer honestly. The answer tells you more about your AI readiness than any technology question.