Before you buy anything, answer these eight questions

Last issue I made a claim that apparently hit a nerve: AI readiness is an operations problem, not a technology decision. A few of you replied to say it re-framed how you were thinking about a project you'd been planning.

So I want to follow through on what I promised — the actual framework I use when I sit down with a manufacturer to figure out whether AI is likely to help them or just cost them money.

There are eight dimensions. None of them are about technology. All of them are things you can assess yourself, today, without a consultant in the room.

Work through these honestly. That last word matters.

Is there a single, trusted source of truth for the metrics that matter to your operation? Or do you have three spreadsheets and two people maintaining their own version of the numbers? AI doesn't resolve data disagreements — it amplifies them. If your team already argues about which number is right, adding a model to the mix makes that argument more expensive.

Is the data that matters actually being collected — and collected digitally? Paper logs, verbal handoffs, and tribal knowledge are real and valuable, but AI can't learn from them. Before you can apply any intelligent system to a process, that process needs to leave a digital trail. If it doesn't, step one isn't AI — it's instrumentation.

AI finds patterns. But if your process changes every quarter — new suppliers, new equipment, new operators — there may not be a stable pattern to find yet. That's not a permanent disqualifier; it means you need to stabilize the process first. Importantly, that stabilization work is worth doing regardless of whether AI ever enters the picture.

Can you name a specific, bounded problem you're trying to solve? "We want to use AI" isn't a problem — it's a solution looking for a use case. The manufacturers who get real ROI from AI started with a crisp problem statement: reduce unplanned downtime on line 3, cut incoming inspection time by 30%, flag purchase orders that historically blow up. Specificity is everything.

Who is accountable for acting on what the AI surfaces? This dimension kills more projects than any technical failure. If the system flags an anomaly and nobody owns the response, the loop never closes. Before you add intelligence to a workflow, make sure that workflow has a human owner who will actually do something with what it tells them.

Do leadership and the floor share the same priorities? AI initiatives that start in the executive suite and get handed to operators rarely stick — and the reverse is equally true. The projects that work are ones where both levels understand the problem being solved and agree it's worth solving. Misalignment here shows up as adoption failure six months after go-live.

Has your team successfully adopted process changes in the past? Not technology changes — process changes. This is a track record question, not a culture buzzword. If the last three operational improvements stalled because people worked around them, that pattern will repeat. Understanding your organization's actual change absorption capacity tells you how fast to move and how much change management to build in.

Is there a real, named owner for this initiative — someone with time, authority, and accountability? Or is it an IT project that nobody on the operations side actually wants? AI implementations fail in proportion to how diffuse the ownership is. This doesn't require a full-time role. It requires one person who cares enough to push it across the finish line and has the standing to do so.

It's not a reason to wait forever. There's no manufacturer who scores strong on all eight before starting. The point isn't perfection — it's informed sequencing. Know your gaps, address the critical ones, then move.

It's also not a vendor checklist. None of these questions are answered by a product. They're answered by honest conversations with the people who run your operation every day. That's both the hard part and the valuable part.

The manufacturers I've seen struggle with AI adoption almost always had a gap in one of these eight areas that they either didn't know about or didn't want to look at directly. The ones who succeeded had usually done some version of this reckoning — formally or not — before they ever signed a contract.

As always — no hype, no vendor pitches. Just what I've seen working in the field.

If one of these eight hit closer to home than the others, I'd like to know which one. Reply and tell me. These issues get better when you push back.

— Chris
Idaho AI Strategies · Boise, ID