Welcome to Manufacturing Minute!
I'm glad you're here.
This week’s episode is a bit different. I definitely went more long form with this one. We are definitely over the “60-second read” metric! But I’ve missed a few weeks, so it evens out.
Let's get to it.
I was tired this afternoon, end of week, long weekend ahead, the kind of tired that's earned and comfortable. Coffee in hand, the warmth from the cup unusually welcome on an unusually cold May afternoon. Catching up with Rob King, who does supply chain analytics and AI at Batesville. He was telling me about his daughter, she's a pretty accomplished high school athlete, with genuine amazement at how strong she'd gotten. This giant of a man with hands that could bend steel, quietly thrilled that his daughter had apparently gotten there first. He said watching her grow like that has a way of asking you to keep up.
Rob is one of those people where you catch the first impression and then realize the mind works exactly the same way. Equally strong, equally never deployed to make a point. Always pushing himself without attaching the expectation to anyone else. You want to learn with Rob not because he lectures but because he's thinking, and if you're close enough you get to watch it happen.
He connected AI to lean the way someone with real lean experience does, not as a framework to cite but as a lens he actually uses. What AI does, done right, is remove the burden of work that doesn't add value and focus the human on what does.
There are three levels of what a workforce can actually do, in lean terms. The first is executing the process correctly and reliably, standard work, the floor performing as designed. The second is improving the process rather than just running it, continuously, as a real cultural practice rather than a bolt-on. The third is rarer: redesigning what the process even is, starting from a different baseline rather than improving the current one. Most organizations implementing lean build the first. Fewer develop the second as genuine habit. The third is uncommon enough that most people who haven't seen it wouldn't recognize it if it happened.
Rob's point was that AI won't move anyone up that ladder automatically. Freeing someone from non-value-add work gets them to the first level reliably. Whether they develop improvement thinking, or the instinct for root redesign, depends on something the technology doesn't determine. Some will use the freed capacity to improve the system. A lot will use it to do more of what they were already doing. Those are very different outcomes, and nothing about the tool guarantees the better one.
His analogy was Excel. Nobody managed that rollout. Operators figured it out independently, in fragmented pockets, without asking permission. Excel in the hands of those operators made them wizards overnight, and the genie never fully got back in the bottle. In lean terms, that's the third kind of change: Excel didn't improve how people worked with numbers, it permanently changed what they expected to be capable of, and the new baseline held. Rob wants AI to do that for systems thinking at Batesville, helping it spread toward genuine capability rather than just faster execution of the same things.
Later in the afternoon, still on the same coffee, which had long since gone cold, I was in a conversation with Mike Carroll and Matt Littlefield. The light had shifted by then, but I was more awake, not less. This is apparently what unwinding looks like for me. Mike has that effect anyway. He carries a breadth that's hard to account for, the kind where you feel like he's lived several lifetimes and taken notes on all of them. He's been preparing to speak in DC on several AI topics, so the thinking was close to the surface. Matt has been building toward the LNS event in Huntington Beach, how an adaptive model can meet organizations where they actually are and free the COO to think about the operating model rather than just run in it. He has a way of seeing the shape of an argument before it's fully formed, and genuinely caring whether it's right, which makes him good to be in a conversation with.
And then Mike brought in education. Adaptive tools that meet students where they are, that free the teacher from the mechanics of differentiated instruction to do what teachers actually want to do, engage, build real relationships, watch a kid figure something out. He's taking that to DC.
Will the teacher know what to do with the time?
It sounds like it should follow automatically, remove the obstacle and the person rises to meet the next level. Except that's not how it works, in a factory or a classroom. You don't climb a hierarchy of needs and stay there. People get pulled back: a floor emergency, a difficult student, a budget crisis, whatever the firefight is that day. The higher levels aren't comfortable places, especially for adults who've found real relief in routine. Thinking about the system rather than executing in it requires sitting with things that don't resolve quickly. Nothing about removing execution burden makes that happen.
A year ago, prompt engineering felt urgent. How you asked the question mattered, and getting good at it was a real skill. That conversation has quieted, not because it wasn't real, but because the next constraint became visible. Now it's architecture: data flows, system design, what triggers what, how outputs in one place become inputs somewhere else without a person in the middle.
AI will get that too. It's already catching architectural mistakes, suggesting better structures, helping people not build the wrong thing. The architecture constraint won't hold as the bottleneck for long.
So what comes after?
If AI handles execution and assists architecture, the remaining question, the one that stays distinctly human, is whether the system should even be doing this, and why, and for whom. Not improving the current system. Redesigning what the baseline is. And the path to more people being capable of that kind of thinking probably looks less like training and more like what Rob described with his daughter: working alongside something that's doing the hard version, watching it, being close enough that you start to keep up. Capacity that develops not from being taught at but from proximity to something worth learning from.
Rob said this afternoon that AI has to start from a position of respect for people, or the whole implementation is wrong from the beginning. Not a philosophical hedge, a design requirement. You don't get people to the level where transformation is possible by removing them from the process. You get there by keeping them close enough to actually develop. If AI done right is something you want to learn with, the way you want to learn with Rob, that changes what we should be building toward.
Not a bad way to go into a long weekend.
How many people on your team are operating at that third level right now? And is what you're building with AI making that number larger, or just making level one faster?
(PS, both conversations happened the same afternoon, but I'm paraphrasing throughout. You get the point.)
Ryan
P.S., if you liked this format, let me know by replying to this email! I may try to start weaving this in periodically.
