Every week I talk to friends and business owners who are convinced they need to "get into AI." They've read the articles. They've watched the demos. They're worried about being left behind.
And almost every time for the business owners, I ask the same question: What does your production data look like right now?
Long pause.
"What do you mean?"
The question everyone's asking — and why it's the wrong one
"Should we adopt AI?" is the wrong question.
It's the wrong question because it frames AI as a decision you make once and then implement. As if readiness is a switch — you flip it, the technology arrives, things improve.
That's not how it works in manufacturing. I've seen companies spend five figures on automation tools that produced nothing because the underlying data was too messy, the processes too inconsistent, or the team too siloed to act on what the system surfaced. The technology worked fine. The operation wasn't ready for what the technology revealed.
The question that actually matters is this: Is my operation clean enough for AI to tell me something useful?
Readiness is an operations problem, not a technology decision
Here's what I've observed across manufacturers of different sizes and industries: the ones who get real value from AI automation share one thing in common before they ever sign a contract or stand up a system.
Their data is trustworthy. Their processes are documented. Their team knows who's responsible for acting on new information.
That's it. No exotic tech stack. No dedicated data science team. Just operational hygiene that happens to be a prerequisite for any intelligent system to function.
Think about it this way: predictive maintenance AI is only as good as your sensor data and your maintenance logs. If technicians are manually entering data in batches, or if shift changes aren't documented consistently, the model learns your documentation habits — not your machine's actual behavior. You've just built a very expensive mirror for your existing inconsistencies.
A useful mental test: If someone handed you a six-month export of your production data right now, could you trust it well enough to make a $50,000 decision? If the answer is "probably not" or "it depends who entered it," that's your AI readiness score. The number isn't zero — but the work to get to "yes" is operations work, not technology work.
The three signals I look for before recommending anything
When I sit down with a manufacturer to evaluate where AI makes sense, I'm not looking at their technology budget. I'm looking at three things.
Data integrity. Is there a single source of truth for the metrics that matter? Or are there three spreadsheets and two people who maintain their own version of the numbers? If it's the latter, any AI layer you add is going to amplify the disagreement, not resolve it.
Process stability. AI finds patterns. But if your process changes every quarter — new suppliers, new equipment, new operators — there may not be a pattern to find yet. Stable, repeatable processes are the foundation. You're not locked out of AI; you may just need to stabilize the operation first, which is worth doing regardless.
Decision ownership. Who acts on the output? This question has killed more AI projects than any technical problem. If the system flags an anomaly or a forecast and nobody is accountable for the follow-through, the loop never closes. Before you add intelligence to a workflow, make sure the workflow has an owner.
What "not ready" actually means
I want to be careful here, because "you're not ready" often sounds like a polite rejection — a consultant's way of deferring a project indefinitely.
That's not what I mean.
"Not ready" is specific. It means: there's a gap between where your operation is now and where it needs to be for AI to deliver on its promise. That gap is identifiable, it's closeable, and closing it often produces value on its own — before a single AI system is deployed.
Cleaning up data entry. Standardizing shift documentation. Assigning clear ownership of key metrics. These aren't glamorous. But they're the unglamorous work that makes the glamorous stuff actually function.
The manufacturers I've seen get the most out of AI are rarely the ones who moved fastest. They're the ones who slowed down long enough to ask — honestly — whether their operation was ready for what the technology would surface.
If this framing resonates with you, I've been developing a simple AI Readiness diagnostic — eight questions that help cut through the noise and tell you where your operation actually stands. No technology agenda. Just a clear-eyed look at the three signals above. I'll have more to share on this in the next issue.
That's it for this issue. As always — no hype, no vendor pitches, just what I've actually seen working (and not working) in the field.
If something here hit close to home, I'd genuinely like to hear it. Reply to this email. I read every one.
— Chris
Idaho AI Strategies · Boise, ID