Most small businesses spending money on AI right now are getting nothing for it. Somewhere between $10,000 and $100,000 into pilots, tools, and consultants that produced no measurable outcome anyone can point to.
It’s not that the technology doesn’t work. It does. I use it every day; my consulting practice runs on it.
The failure isn’t technical. It’s that most businesses skip the one question that would tell them whether they’re ready to buy anything at all.
Here’s the question: what specific process in your business, that you can name in one sentence, is costing you time or money every week that AI could actually help with?
If you can’t answer that question, you’re not ready to buy AI tools. You’re ready to run an inventory of your own manual work — and until you do that, every AI investment is a guess with your operating budget.
The inventory before the technology
Here’s the pattern I’ve watched repeat across dozens of engagements: the businesses that get real value from AI adoption always start in the same place. Not with the technology. With their own operations.
An operations director I worked with last year was spending three hours every Monday morning assembling a weekly performance report by copying numbers out of four different systems. At her fully-loaded cost, that report was consuming $3,700 per year in her time alone — before you count the delay, the transcription errors, and the fact that she couldn’t do anything else while she was doing it.
We automated it in two working days. Cost under $2,000, delivered. The report started going out at 7 a.m. Monday instead of 10 a.m., with fewer errors and none of her time.
That’s not the interesting part.
The interesting part is what happened next. Within a month, her team had identified six more processes that fit the same shape. Nobody had to sell them on AI adoption anymore. They watched one boring, useful win happen, and they suddenly saw where else the pattern fit.
The lesson generalizes: AI adoption stalls not on technology but on pattern-visibility. Once your team sees one example of the shape, they’ll find the rest for you. Before that, no vendor pitch will land.
What “commitment” actually means
When people say AI adoption requires commitment, they usually mean something vague and inspirational. Here’s what it actually means, practically.
Commitment means someone in your business writes down every recurring manual task that takes more than one hour of human time per week. It might take two weeks of asking people what they actually spend their time on. It will be uncomfortable — you’ll discover work no one told you about, and processes people invented to route around broken systems. That inventory is worth more than any AI tool you can buy, because it’s the map of where AI would actually help.
Commitment means you’re willing to kill projects that don’t work. Most AI pilots fail quietly, and the ones that fail loudly are the ones nobody was willing to stop. If your culture can’t say “this didn’t work, we’re moving on,” AI adoption will multiply your existing pattern of accumulating half-finished initiatives instead of solving anything.
Commitment means you’ll measure honestly. Baseline what things cost now — in hours, dollars, or delay — before you touch anything. Compare against reality afterward. Don’t let the vendor case studies do your measurement for you. A “30% productivity improvement” is meaningless without your baseline; some improvements you’d measure yourself would be smaller, but real, and durable in a way the vendor number never was.
The honest assessment
None of this is exotic. None of it requires a data science team. What it requires is a leader willing to spend a quarter running the discovery work before writing a purchase order.
That’s the actual bottleneck for most small and mid-sized businesses. The tools are cheap. The consultants are available. The willingness to slow down before speeding up is what’s rare.
I built a free tool that helps with exactly this problem — the AI Readiness Diagnostic. It’s ten questions, takes about ten minutes, and its most useful feature is that a low score is actually good news. A low score means you know where you stand. Most organizations don’t.
The scoring bands are honest: if you land in the lowest band, the recommended action is not to launch a formal AI program yet. Build the manual-process inventory first. Publish a simple AI usage policy. Baseline one team’s cycle time. Then come back in six months.
That’s not the advice most AI vendors give small businesses. It’s the advice that actually works.
Start here
You can download the diagnostic at derivativeresearchsystems.com/diagnostic. No newsletter enrollment, no consultation gate, no dark patterns. Use it internally, share it with your team, adapt it to your context. It’s licensed CC BY 4.0 — attribution is appreciated, not required.
If your score surprises you — high or low — sit with it for 24 hours before acting on it. The questions you scored zero or one on are the highest-leverage places to start. That’s where the honest work is.
AI adoption is within reach for your business. Not because the technology is easy, but because the questions are answerable if you’re willing to ask them honestly. The businesses that will get real value out of this moment are the ones running the boring, foundational inventory now — while their competitors are still watching demos.
The demo is the easy part. The inventory is the work.
Jeff Crump is the founder of Derivative Research Systems, an AI adoption advisory practice for mid-market businesses. He has 20+ years in production financial systems and previously held enterprise architect roles at Rockwell Automation, TIP Technologies, and Harley-Davidson.