I recently joined ASTM International's F50 committee, the group writing the standards and
methodologies for putting AI to work in manufacturing. I'll be upfront about where I'm
coming from. I didn't arrive at this as an AI researcher. I came to it from the additive side,
where I've spent years watching a technology go from interesting demo to qualified
production part. That climb is the reason I wanted a seat at this table, because I think AI in
manufacturing is about to make the same climb, and for the same reason.
The technology is moving fast. That part isn't in question. The harder problem, and the one
that hasn't really changed, is turning AI from something with potential into something that
actually changes how a shop runs.
The clearest way I've heard that problem framed comes from Dr. Jay Lee of the University
of Maryland, who chairs F50. In a recent paper, he and his colleagues lay out what they
call the ABCDE of engineering AI, the five ingredients every system depends on:
-
A Algorithms: the models and methods that power an AI system.
-
B Big Data: the information coming off machines, processes, and operations.
-
C Computing: the chips and infrastructure to process and scale it.
-
D Domain Knowledge: a real understanding of how the process and the parts behave.
-
E Ecosystem: the people, partners, integrations, and standards that make it work in practice.
Here's the part that matters. The first three, the algorithms, the data, and the compute, are
increasingly available to everyone. You can rent the compute, download the models, and
collect the data. They are becoming commodities. What you can't download is domain
knowledge and a working ecosystem. Those are the real differentiators, and they happen
to be the two things a manufacturer with real operational experience already has. The
shops that get results from AI are the ones who ground it in how their process actually
works, not the ones chasing the flashiest model.
That is also where standards come in, and it is why I think they matter more than the tools
do. The moment you try to deploy AI on a real line, the same questions surface every time:
whether the system actually works, whether the data feeding it is good enough to trust,
what success looks like and how you measure it, and how any of it connects to the
systems you already run. None of those are questions a single vendor's product can
answer. They are industry questions, and the only way to answer them consistently is with
shared standards. Standards are what reduce the risk, let different systems work together,
and build enough trust in the results that people will actually put them into production.
Without them, every AI project stays a one-off, and one-offs don't scale.
This is familiar ground for me, because it is exactly what happened with additive. The thing
that moved 3D printing off the prototype bench and into real production wasn't a better
printer. It was the standards work that let an aerospace or medical buyer trust a printed
part. ASTM's additive committee, F42, governs how a part is physically made and qualified.
F50 governs the data and AI layer that runs on top of it. Same factory, two floors. If you
already work on the first floor, the second one is the next step up.
So I see this as a natural progression for any manufacturer already investing in digital
processes, additive, 3D scanning, and connected workflows. The value of AI is tied directly
to the quality and availability of your data, and shops that have already gone digital have
that foundation in place. They are often the best positioned to use AI, because the hard
part, getting good data off the floor, is already underway.
AI is going to keep advancing whether any of us are ready or not. But whether a
manufacturer gets real value from it or just ends up with a graveyard of stalled pilots will
come down to more than tools. It will come down to practical frameworks, shared
standards, and a willingness to work across the industry instead of in isolation. That is the
work F50 is doing, and it is why I joined.
At M5D, we are putting real effort into shaping how AI gets applied on the factory floor, and
into helping our customers figure out where it delivers and where it doesn't. If any of this is
on your radar, the F50 committee is open, and it could use more people who know what
the floor actually looks like. Take a look, and consider getting involved.
Learn more and sign up for the ASTM: ASTM International Forms Artificial Intelligence Committee for Manufacturing | ASTM
