I was chatting with the CFO of a manufacturing company recently about AI and where it could actually make a financial difference.
I asked him where he saw the biggest opportunity.
He said:
“I closely monitor P&L savings. But I think an even bigger impact AI can make is on CapEx — helping us make better decisions about what we invest in and whether we really need to invest in it.”
That got me thinking.
We spend a lot of time talking about reducing inventory, getting better supplier prices, standardizing parts and so on.
But what happens when a company is about to spend money on a new machine? Or a die, fixture, tool or gauge?
Say I'm looking at a proposal for a new die.
The first thing I'd want to know is:
Do we already have something that could do the job?
Maybe we made a similar component five years ago. Maybe it was for a completely different product. Maybe there is already a die sitting somewhere that could be used or modified.
The trouble is, finding it isn't easy.
The part number could be different. The geometry could have changed. So could the material or tolerances. It may even have been made at another plant.
This is where AI starts getting useful.
Take the new component and look across what the company has made before.
Have we made something like this before?
If yes, how did we make it? What tooling did we use? Which machine? Do we still have it?
Then the engineers can figure out whether any of it is actually suitable for the new requirement.
Now let's say they find an old die and tell me it can be modified.
Great.
But I'd have another question.
Have we done this before?
Maybe my predecessor faced the same choice five years ago. Maybe another plant did.
What did they do?
Maybe they modified the old die, saved the money and it worked perfectly.
Maybe they tried exactly that, had problems during ramp-up and ended up buying a new die six months later.
I'd definitely want to know.
What did they decide? Why? How much did it eventually cost? Did it delay the program? Were there quality issues?
Because all of that should have some bearing on what we decide today.
Manufacturing companies have decades of this experience.
Machines bought. Dies modified. Suppliers changed. Parts redesigned. Projects that went well. Projects that didn't.
The information is probably there too — in drawings, CAD files, purchase orders, engineering changes, quality reports, project documents and people's heads.
The problem is connecting it when you actually need it.
I think AI could become very useful here.
Imagine sitting down to review a new CapEx proposal and being able to ask:
Have we faced something like this before? What did we do? And what happened?
You might still approve the new machine.
You might discover that something you already own will do the job.
Either way, you're making the decision knowing a little more about what your company has already been through.
And that seems like a pretty good place for AI to help.