I used to have a guy on my team named Greg. Surface finish was his thing, and he had forty years of it in his head. A good chunk of his week went to answering questions. Hey Greg, how do I do this. Hey Greg, why is this part coming out like that. Everybody went to Greg, because Greg knew.
I think about him all the time now. If I could have taken what was in Greg’s head and built an “Ask Greg” for surface finish, I would not have replaced Greg. I would have freed him up to mentor the young engineers and teach them everything that was never written down.
That is the whole idea I want to leave with you. A lot of the jobs sitting on your org chart were designed for a pre-AI world, and most of us have never gone back and looked at them since. So before you approve the next requisition, look at the work through an AI lens.
I came up from the dish room at Ford, through a tool and die apprenticeship into engineering, and then into global manufacturing leadership at Ford and American Axle. I have watched a lot of technology come through a plant. I have never seen anything change the design of work this fast.
Most of the jobs on your org chart were built for a different era
Take training. Every company has to train. New people come in, we have to get them up to speed, and we want to up-level the people already there. The old answer was a team of trainers.
The new answer looks different. You have one person who holds the standard for how training gets done. How we write our single point lessons. How we write our work instructions. How we build the material. The AI does the production, and your person brings the judgment about what training needs to happen in the first place.
That is what reorganizing means to me. Pull up the org chart and ask which tasks, which processes, and which outcomes can shift over to AI, and then ask how you elevate the people doing them today.
I am not talking about getting rid of anybody. In manufacturing, finding talented people is hard enough as it is. I am talking about moving them toward mentorship, innovation, and growth, which is where their judgment was always worth the most.
Things are moving fast, and now we have agentic AI on top of it. Take a step back and think about what the work looks like in one year, three years, and five years, then redesign around that.
Start at home, because that is where the AI glasses come from
There is a lot of backlash to AI right now, and plenty of people are afraid of it. Maybe they used it once for an email, and it lied to them. That happens when you use it poorly. So I tell people to start at home, where the stakes are zero.
Here is my example. I moved recently and consolidated two homes, and for the first time in my life I got organized. All the pens in one place. All the highlighters in one place. And while everything was fresh in my mind, I decided to build myself a custom GPT that could find anything in the house.
I opened ChatGPT, put it in voice mode, and walked around narrating. Here are the pens. Here are the batteries. Every room, every drawer, every bin, the garage, the crawl space, all of it. ChatGPT turned it into a searchable spreadsheet, which I never had to build myself, and I put that on the back end of a custom GPT.
Now I open it up and ask where that soup pan went, and it tells me. I named it “Where Did I Put That?” It gets used more than you would think.
Once you do something like that at home, your brain starts asking a better question at work. A manufacturer with no fancy enterprise inventory system could run a tool crib the same way. Small shop, real problem, solved in an afternoon.
The more you use it, the more it rewires how you see things. You hit a problem, and your first instinct becomes how could AI help me with this. That is what I mean by walking around with AI glasses on. Start with your to-do list and look for what drains your energy, what everybody hates doing, and what repeats.
What the lens shows you once you walk the operation
When smaller manufacturers tell me AI is a technology thing for the computer team to handle, I point them at the people side of the business, because that is where the fastest wins are.
Training again, since a lot of companies sell training. Get that knowledge out of one person’s head, teach several people to deliver it, and now you are looking at additional training you could package and sell. That is a revenue conversation, not a cost conversation.
Onboarding is another one, and turnover in manufacturing keeps that door swinging. Build an AI that carries the whole thing: what the new hire does in the first 30, 60, and 90 days, what the buddy does, what the hiring manager owns, what HR owns, and the early signs that somebody is struggling.
Knowledge capture is the Greg story. I also had a boss named Phil, and I knew exactly how Phil thought. Before any presentation, I asked myself the questions I knew Phil would ask. Other people would walk in and get flattened by a question they should have seen coming. Today I would build a small AI and ask it what Phil would say and what Phil would push on.
Then there is process. We have to have process, or we get quality problems. Narrate how the work gets done on the floor, let AI write it up, and when an engineering change comes through, run the change against the process and ask what should concern you. AI is good at patterns. Load a workspace with your lessons learned and your past FMEAs, and it will catch things a tired human misses at four o’clock on a Friday.
Governance and use cases have to move at the same time
Most companies start with a governance committee, and I think that is right. Find the early adopters, the people already interested, and put them on it.
You need the policy and the guardrails in place so customer data stays protected and people know what belongs in a prompt and what never does. Without that, somebody is hiding in the bathroom running company information through ChatGPT on a personal phone. That is the risk you are carrying while you decide.
Governance on its own is not motivating, though. The use cases are what get people excited, and that excitement is what pulls them through the unglamorous work of writing the policy. Run both together. Start pilots, get early wins, talk openly about what worked and what did not, and then ask how you replicate it across similar work.
Cost deserves its own conversation. Nobody knows whether these models stay subsidized at today’s prices. My favorite is Claude, I also use ChatGPT, and most of my clients are on Microsoft Copilot, so I have to know all of them. One of those tools moved to usage-based pricing on me, and my bill went up. A leader I spoke with is staring at a $500,000 hit to his cost center once you add up the licenses.
Get people from different functions in the room, and protect the psychological safety in it. People need to be able to push back, ask the dumb question, and play devil’s advocate. You want all the voices at that table.
I manage AI employees now, and your team will too
I am a solopreneur, and I have a whole staff of them.
My marketing workspace is my copywriter. It helps with LinkedIn posts and the newsletter, and I built a skill that runs every Monday morning searching for speaking opportunities. I used to pay a person to do that. Another skill reads through my transcripts and suggests what my next newsletter should cover.
Getting there took work. I loaded the back end with my website copy, my testimonials, my top posts, my values, and my about page. Then I asked for ten blog ideas, picked one, asked for 300 words, and went back and forth telling it I would never say that, until I could ask how to rewrite the custom instructions to get the right tone next time.
My sales workspace reviews my call transcripts and tells me where I could have done better, and it is loaded with the objections I hear. My leadership brain holds my team’s personality profiles, the StrengthsFinder and Kolbe and Myers-Briggs results, so when a hard conversation is coming I can think it through first. One time it told me exactly what I needed to say, and then told me I probably would not say it, because I avoid conflict. It was right. My financial coach knows my money story and my P&L.
For a manufacturer, think about the person who has to quote a job. A workspace loaded with past estimates, past proposals, and the pricing that won gives them a real starting point the moment a new RFQ lands. Keep a human in the loop, because judgment still belongs to us. But all that manual, non value added work in the middle can move.
One warning from my own experience. When I first got into this I had six or seven things running at once and my brain could not keep up, so I had to slow myself down. I was filling every hour I freed up with more work. Decide in advance what your people will do with the time they get back, or the time will get eaten and nobody will feel the benefit.
So do you make the hire
It depends on the situation, and I would never tell you otherwise without knowing yours. What I will say is that the AI strategy should come first, so you know which tasks, outcomes, and processes are moving before you write the job description.
Say you need a curriculum designer. Do you need that hire, or do you need a strategist who can make the curriculum designer you build inside AI much stronger.
I worked with a company whose power lockout training was ten years old. They advertised it as a four hour course and it took them two and a half hours to deliver, so the mismatch was already costing them credibility. We modernized it, brought it up to current standards, and made it substantial enough to match what they were selling.
Once they saw that, they started listing other training they could build and sell. And because they do power lockout, they have ECPL placards to maintain, so the next idea was an AI checker that reviews every placard before a human does. People miss things. The AI catches the obvious ones and gives the human a better place to start.
The question underneath all of it is simple. Look at how you are structured, ask which of these things AI can carry, and then ask how you scale up the team you already have so they can see the same opportunities you do.
Your people are already watching
I see companies in three buckets.
The ones with governance, training, and use cases in place, who are up-leveling their people and tying prompts, agents, and skills back to ROI
The ones in the middle, working through governance and cost, figuring out where AI belongs and where it does not
The ones who have not started and have everything locked down, which is the riskiest spot, because people will use it underground anyway
Whatever bucket you are in, your team has questions they are not asking out loud. Somebody is thinking that training the AI is training their replacement.
The genie is out of the bottle on this one. I believe AI is going to be the biggest career differentiator any of us have seen, so I tell people to take the learning while it is being offered. The tasks were never where the value was. The value is in the judgment, and AI gets you to the judgment faster. When somebody tells me AI is going to take their job, I tell them a person who knows AI is more likely to take it.
Train people before you roll anything out. Let them understand the risks and the limits, and then give them somewhere to talk about it, because the team conversation about what broke and what failed is the part companies skip. I say if you are not ready to throw your computer out the window, you are not trying hard enough.
The other mistake I watch companies make is piling on more work because AI is supposed to make it easy now. People are getting overloaded. Listen to them.
Where to start on Monday morning
Start with your biggest problem. Not the tool, not the platform, the pain.
Say scrap is your problem. Here is how I would set it up:
Give it a persona. “You are a quality expert. You have solved quality problems and saved companies millions of dollars over the years.”
Give it the context. “Our scrap was 1 percent and it has gone to 5 percent. We have looked at this, this, and this. We cannot figure out what is going on.”
Give it the request, then ask for pushback. “Give me three things I can go look at and change today that will drop my scrap. Then ask me any clarifying questions.”
Context is the whole game. I do not like to type, so I put it in voice mode and talk, and even when I ramble it makes sense of what I am after.
If you want a smaller first step than that, open it up and say: this is my company, this is my role, give me three unfair advantages AI could give me here. Then get a paid account for twenty dollars a month, go into settings, and turn off using your data to train the model. Protect yourself while you learn.
Leadership is still the skill. Your energy, your judgment, how you show up, the relationships you hold together, all of that stays human, and it is the front end of everything. AI is the back end. Do not let go of the human part while you are busy building the rest, and do not let anybody on your team get left behind in the age of AI.
If you have an open requisition sitting on your desk right now, tell me in the comments what the role is. I will tell you which pieces of it I would run through the lens first.
I’m Cheryl Thompson. I came up from the floor at Ford through tool and die into global manufacturing leadership at Ford and American Axle, and I now teach manufacturers and their teams to put AI into real operations. I run a free Learning Lab most weeks, 60 minutes, immediately applicable. If you want the next one, you can see the schedule on my website.