June 12, 2025 • 33 mins
As AI adoption accelerates across the supply chain, many manufacturers are realizing that success requires more than just technology. Scaling AI is not only about building better models: it also depends on strong data infrastructure, effective training programs, and capturing institutional knowledge before it walks out the door.
This week, host Ellen Wood speaks with Bryan DeBois, Director of Industrial AI at RoviSys, about the deeper challenges of implementing AI in industrial operations. Bryan shares insights from more than two decades in the field, offering a practical look at what it takes to move from isolated pilots to scalable, long-term solutions. He also explores how autonomous AI is beginning to reshape the future of manufacturing and why the human element still matters most.
In this episode
This week, host Ellen Wood speaks with Bryan DeBois, Director of Industrial AI at RoviSys, about the deeper challenges of implementing AI in industrial operations. Bryan shares insights from more than two decades in the field, offering a practical look at what it takes to move from isolated pilots to scalable, long-term solutions. He also explores how autonomous AI is beginning to reshape the future of manufacturing and why the human element still matters most.
In this episode
- Why industrial supply chains struggle to scale AI successfully
- The role of data maturity, training, and expert knowledge transfer
- How AI can drive both operational excellence and sustainability
- Why autonomous AI is the next leap for manufacturing leaders
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:06):
You're listening to Speaking of Supply Chain, a meeboch podcast.
This is a show for logistics professionals looking to learn
more about the latest innovations in supply chain. Each episode
will feature a conversation on topics such as mitigating supply
chain disruption and reducing risk, current automation trends, sustainability initiatives,
and more. Let's dive right in.
Speaker 2 (00:30):
Hello and welcome to Speaking of Supply Chain, where we
explore trends, current events, and innovations impacting the logistics and
supply chain industries. I'm your host, Ellen Wood. AI is
rapidly transforming manufacturing by enhancing efficiency, reducing costs, and addressing
labor shortages. Recent stats indicate over eighty percent of businesses
(00:50):
have adopted AI to some extent, with thirty five percent
implementing it across multiple departments. But one area that's still
working to catch up is in dust operations. From predictive
maintenance to automating repetitive tasks, the industrial manufacturing sector is
rapidly finding ways to incorporate these advanced capabilities. Joining me
(01:11):
today to dig into some of the details is Brian boy,
director of Industrial AI at Rovisus, welcome Brian.
Speaker 3 (01:18):
Hi, thanks for having me.
Speaker 2 (01:20):
Great to have you. So before we get into our
discussion today about AI, let's take a moment just to
get to know each other a little bit better. This
is a feature that a lot of our listeners are
probably not used to. It used to only be available
in our video content, but now that video is available
on Spotify, we're putting it into the full program. So
(01:43):
our icebreaker question for today, I've got one for you,
and then you've got a different one for me. So
mine for you. What's a weird fact about you?
Speaker 4 (01:53):
So I would say the I live a pretty normal
suburban life. But I would say the one weird fact
is that my wife and I are kind of roller
coaster freaks, and so we both have each ridden I
would say at least over one hundred rides on roller coasters,
probably one hundred and fifty at this.
Speaker 3 (02:11):
Point when we met.
Speaker 4 (02:13):
We met at eighteen in college and both kind of
bonded over roller coasters. We live in Ohio and so
we've got Cedar Point is up the street from us,
so that's you know, that's kind of why. And then
when we raised our kids, we kind of raised them
to be roller coaster freaks, so they ride with us now,
so all five of us jump on these coasters.
Speaker 3 (02:32):
But yeah, we love them. We love them big and
fast and scary.
Speaker 2 (02:36):
So my kids also love roller coasters, and my husband
is a roller coaster freak. We've been to Cedar Point
multiple times. There's one roller coaster there. I forget which
one it is. It's the ginormous one that's so tall
that you can see Canada on a clear day.
Speaker 3 (02:50):
Yeah, that's money and force.
Speaker 2 (02:52):
The millennium force. So I remember, I love riding that
roller coaster. But I read something recently about it that
was really interesting, and that is that it is one
of the few roller coasters in the world that can
help break up a person's kidney stones. Totally random. If
you have kidney stones, you can ride that roller coaster
and it will help break them up so that you
(03:13):
can pass them.
Speaker 3 (03:14):
I don't know.
Speaker 4 (03:16):
I don't know how accurate that is, but uh, I
guess as I age, that's good to know.
Speaker 3 (03:20):
If it's true.
Speaker 2 (03:22):
That speaks to the force of that roller coaster.
Speaker 4 (03:24):
It is the first hill. It's wild, it's wild.
Speaker 2 (03:28):
Yeah, yeah, yeah. I was talking about roller coasters just
this past weekend with my son. We were driving through
Louisville and looking over and seeing when Kentucky Kingdom was
going to open, because it's just off I sixty five
in Louisville. You can see it right by the fairgrounds.
And he's he's very excited. He's like, when is it
going to open? Can we go find roller coasters?
Speaker 3 (03:49):
Yep?
Speaker 4 (03:50):
Yeah, I can't wait. We got our season passes. I
can't wait for see your point open.
Speaker 3 (03:54):
We'll be there.
Speaker 2 (03:55):
So all right, that sounds like a lot of fun
for the summer. So you had a different question for
me to get to know me a little bit better
for our listeners.
Speaker 3 (04:03):
And what was that I did? So, Ellen, I like
to cook a lot.
Speaker 4 (04:07):
I do all the cooking in our house, so I'm
always looking for new ideas, new inspiration.
Speaker 3 (04:11):
What's your favorite dish to cook?
Speaker 2 (04:13):
Oh? My favorite or the one I cook the most often,
because that's a different thing all together.
Speaker 3 (04:19):
I would say, what's your Yeah, what's your favorite?
Speaker 2 (04:22):
My favorite? So I love making soup, and yeah, I
make soup all the time. And I've I've worked this
past winter and fall at perfecting my bone broth. Okay,
and I've been making multiple batches of it and I
make bone broth at least once a week, and I
(04:44):
freeze it. I give it away to friends. But I
just I've wanted to find like the best bone broth,
and I like beef bone broth, and I have I
have a grocery store near us. It's an international grocery
and it has a lot of different ethnic that most
grocery stores do not have. And this store has all
(05:04):
the parts that you need for making good broth that
most grocery stores aren't going to have prepackaged nicely for
someone to go consume. And one of the biggest additions
to my chicken broth is the feet, oh chicken.
Speaker 3 (05:19):
Okay, okay.
Speaker 2 (05:22):
You get so much more collagen out of chicken feet,
and it's not something you will ever find at a
regular grocery store. But if you can get your hands
on them from a butcher and add them into your
your bones and your your vegetables and your herbals that
you put in there, that the chicken feet make all
the difference in a very very collagen rich broth, which
(05:43):
is what you want.
Speaker 3 (05:43):
Yeah, well, that's a good tip. I like that.
Speaker 2 (05:46):
It makes my kids laugh every time I make it
because there's these little chicken feet stickn open.
Speaker 3 (05:56):
Well, you know, as long as it tastes good in the.
Speaker 2 (05:58):
End, exactly, it hasn't been them from eating it. So
all right, Well, let's get on to our topic today,
which is not chicken feed or roller coasters. It is
AI in industrial manufacturing. And this is something that you
have a passion for, So tell us a little bit
about it.
Speaker 4 (06:17):
Yeah, So, well, I mean to get there, I kind
of have to start back at my journey here, my
career at Robusis. So I work for a company. So
we're a system integrator and we target manufacturing and industrial customers.
So I started here actually right out of college. I
co opt here and then I started full time. So
I've actually spent twenty five years here at Robusis has
(06:38):
been a health professional career, which you don't see as
often any more, folks spending kind of their whole career
at a single company. And I came in with a
software background, and so at the beginning of my career,
we were writing a lot of custom software and things
like that for customers. Now, things that progress, there's a
lot of commercial off the shelf software that you can buy,
(06:59):
so so in some ways we've kind of advanced and
now so we're focused on AI for those same.
Speaker 3 (07:05):
Industrial and manufacturing customers.
Speaker 4 (07:07):
I'm very passionate about it because I you know, I've
spent I really came into this industry knowing nothing about manufacturing.
Speaker 3 (07:13):
When I started twenty five years ago, I.
Speaker 4 (07:15):
Knew I liked the show How It's Made, so that
was kind of cool, like to seeing how stuff was made.
And so now I've spent the last two and a
half decades working in this industry and really have fallen
in love with manufacturing. I do feel like manufacturing is
the biggest lever that we can pull as a civilization
to create the most prosperity for the most people. I
(07:37):
think it's that important. I think it is the wealth
generation mechanism for for civilization, for societies, certainly for economies,
and so I'm.
Speaker 3 (07:47):
Very passionate about it. I'm passionate about what we do.
Speaker 4 (07:49):
And manufacturers in general are not known for being real
cutting edge right there for a lot of good reasons.
They're very risk averse, and so they're typically five to
seven years behind more of the IT world and in
the rest of the world other industries, and so it's
been you know, a longer process to get them to
(08:12):
adopt AI. But all of I mean almost all of
my customers now are taking some kind of steps towards that.
Even if they haven't gone full born AI yet, they've
got a data science team where they're at least starting
to build out some of the infrastructure, do some of
the infrastructure upgrades that are required to get to an
AI future?
Speaker 2 (08:28):
All right, So once that initial AI project gets off
the ground, how can organizations effectively scale that up and
make it into those bigger projects that don't necessarily overwhelm
the team As you said, they're they're risk averse. They're not.
This isn't their wheelhouse, this isn't where they look for
(08:50):
the next thing, and this is this is very new.
So how do we do it without overwhelming those teams
or risking the quality of that implementation and try to
do something We certainly don't want to do something and
fail because then you know, it's it's going to make
the team risk averse to trying something new again.
Speaker 3 (09:08):
Right, Yeah for sure?
Speaker 4 (09:10):
And I know it sounds self serving, but the answer
really is to hire a system integrator. So that's that's
really our role is to in cases where the customer
doesn't have any internal capability like that, any internal team.
We're there to bring in the experts and to be
the horsepower behind those types of projects. But when we
also have plenty of customers that have existing you know,
(09:33):
data science or AI teams and then we come in
to scale their capabilities out. So you know, it's like, look,
we want to you know, we've heard this where we've
got some models built and we think we were onto
something here, but we want to roll this out to
five sites. So we've got a three year plan to
roll us out to five sites. I'm like, wait, I
can scale horizontally, I can bring in bodies and we
(09:55):
can do we can roll it out to five sites
in parallel. And so that suddenly is is an eye
opener there when they realize we.
Speaker 3 (10:04):
Could shrink this schedule a heck.
Speaker 4 (10:05):
Of a lot more So, like I said, I know
it's sound self serving, but that really is the role
that system integrators play in these types of things is
to augment that capability.
Speaker 3 (10:14):
I will say.
Speaker 4 (10:14):
Longer term though, one of our key tenants at robsis
is that we believe that customers should own the solution.
Speaker 3 (10:22):
When we're done. So we spend a lot of time
on training.
Speaker 4 (10:26):
You know, we tell customers you're allowed to look over
our shoulder the entire project if you'd like. We want
to make sure that you feel comfortable owning this system
when when we're done, because I mean, it is your
system ultimately, and that's that you know. So we you know,
we know that they're going to call us when they
have the next big project or whatever. But we don't
want to lock them in into a situation where, you know,
every time they have to retrain the model or add
(10:49):
a little you know, add some tags to it or something,
we don't have to be able to force to call
us for that kind of thing.
Speaker 2 (10:56):
Sure, so as it becomes more embedded in these workforces,
as you're handing off the capabilities and the tools to
do it, what new skills should companies be proactively you know,
teaching their employees and getting people to a comfort level
with this type of technology, and how can they best
(11:17):
approach the training for those things? Because I know we've
had a number of trainings in our organization on AI.
We had a workshop just yesterday trying to you know,
workshop and get some new ideas, some fresh ideas that
aren't just what's your chat GPT prompt? I mean, that's
that's the one I use the most. But how do
(11:38):
we train this next generation of workers to function with
the AI and not fear it.
Speaker 4 (11:47):
Yeah, so there's a couple of different answers there, because
it really does depend on what your role is in
the organization. So if you're somebody who actually feasibly you're
an analyst or an engineer, and you feasibly would be
maybe interfacing to these models or maybe even building these models,
you're going to need some kind of data science background. Now,
(12:07):
the good news is that there are free resources. There
are in resources available, so there's lots of ways to
get some of that data science background. I will say though,
that that doesn't seem to be as much of a
problem for our customers where they're seeing the bigger problem
is in the loss of expertise. So we are seeing
(12:28):
some people are calling it a silver tsunami, but we
are seeing a loss of primarily baby boomers post COVID
who are leaving the manufacturing workforce, and what we are
not seeing is a corresponding increase in younger folks entering
the manufacturing workforce. So just to give you some SPAT stats.
This came from LNS Research in twenty nineteen. The average
(12:51):
tenure in the this is for US. The average tenure
for the manufacturing workforce average years at a company is
twenty years. The average tenure, So that was twenty nineteen
and twenty twenty three, that average tenure fell to three years. Yeah,
the average time and position fell from seven years to
(13:12):
nine months in twenty twenty three, in just four years.
It's a significant issue that we're seeing. It is the
number one issue that customers bring up when we sit
down with them nowadays. It's a continuing issue. And so
there's two aspects of that then that we typically talk about.
One of those things is how can we build AI
(13:33):
systems that can augment the capabilities of a novice person
someone who hasn't been at the place very long. They
also have very high turnover in these positions, so oftentimes
you're talking about operators on the line. You know, you
had a ten to fifteen year person and the person
replacing them maybe it has six months to a year
(13:53):
in that position. Because the high turnover, they're constantly starting
over again. So how can we build AI systems that'll
supplement the capabilities of these no office operators so that
we're not just running these lines suboptimally forever. And the
other aspect of it is capturing that expertise from their
subject matter experts before they leave. So I was just
(14:16):
at a conference. I was speaking at a conference this
week and that was one of the key things they said.
Somebody raised their hands said, you know, we've got We've
got some of those people that you're talking about that
have been here for fifteen twenty years, but they are
on their way out. I mean within the next couple
of years they're playing and retiring. What can we do
to capture that expertise in AI? And the good news
is is that there are mechanisms to mechanisms to do that,
(14:38):
but the time to do that is now, before you
lose that expertise out the door.
Speaker 2 (14:42):
Yeah, that really has to be a proactive program to
get that information while there's still the opportunity to do so.
And it's great that they're AI tools to do that
as well. But one of the things you were talking
about earlier was, you know, driving the efficiency and the
productivity within that manufacturing facility. How can they also leverage
(15:08):
that AI to advance sustainability goals or something like energy
management or waste reduction. You know, there are a lot
of other things besides, just like you said, the manufacturing
process itself not that excited about changing the way they
do it. They're five to seven years behind. So there
(15:31):
are other ways that they can use this technology to
improve operations. That isn't just the manufacturing process, but more
about the facility itself and the processes of or I
guess the byproducts of the processes. There's so much involved
in manufacturing and AI is going to touch all of it.
Speaker 4 (15:50):
Yeah, yeah, And so it's actually it's so interesting to
kind of generationally to see because especially when you talk
to millennials, when you talk to gen z's sustainabilities top
of mind for them, and many of them are now
in leadership positions at these companies, and so that's good.
Like so, so we want to be cognizantve sustainability. There's
(16:11):
kind of two aspects there. One is that we need
to start to define AI projects as sustainability projects, because
you mentioned a whole bunch of things, and every single
one of those things. If I'm increasing the efficiency of
the line, I am implicitly I am reducing energy. I
am I am increasing throughput, which means that I'm reducing
waste all of those things. I mean, the amount of
(16:34):
waste in manufacturing is mind bibling when you see how
much gets thrown out, how much draft gets thrown out, uh,
and then the amount of energy that's used to run
these facilities. So if I'm doing an AI project and
I'm squeezing you know, singlar double digit percentage improvements out
(16:55):
of those lines, I am inevitably reducing the amount of energy.
Speaker 3 (16:59):
You know.
Speaker 4 (17:00):
We did an AI project for a glass manufacturer, and
it's so funny too, because depending on who you talk
to at the company, you tend to get a different answer.
So we did a glass project and we were building
an AI optimizer to more efficiently cut glass. Think like
window panes out of a big sheet of glass. Obviously,
everything else that's not window is scrap. It gets broken
(17:20):
down and it gets remelted. Now, we were talking to
a couple of the operators and they go, well, we
don't have scrap because all of that gets that gets remelted, right, So,
so effectively our scrap level is zero. Then later on
in the day, you have a meeting with one of
the finance people and they're like, no, all of that
is scrap. We track all of that, even if they
don't know, we track all of that because all of that,
(17:42):
even the melting down. You think about all of the
energy that went into producing that pane of glass the
first time, the first time. Now you're gonna break it down,
you're gonna smash it, and you're gonna remelt it and
pour all of that energy and raw material back into it,
all the labor that it took. Absolutely so all of
that is wasted energy. So we were made able to
make massive Their optimization engine that they were using to
(18:04):
lay out these glass panes was very suboptimal. We were
able to squeeze so many more panes of glass out
of out of these sheets of glass and reduce their
scrap by double digit percentage. That's a huge impact on
energy reduction. Was that classified as a sustainability project by
corporate It was not.
Speaker 3 (18:20):
It should have been.
Speaker 4 (18:21):
So that's what I'm saying, we need to start defining
these AI projects as sustainability projects.
Speaker 3 (18:26):
But the other aspect of it is.
Speaker 4 (18:28):
That I think is in some ways even more important
when we look at sustainability goals, is many of these
companies have very aggressive sustainability goals. Great, we want to
be carbon neutral bi X here. Fine, here's the problem
with that is there is no other technology other than
AI that is like in the wings waiting to reduce.
Speaker 3 (18:48):
Their greenhouse impact or their carbon like.
Speaker 4 (18:51):
There isn't any other technologies other than AI. So if
you want to get hit these aggressive state sustainability goals,
which again I encourag that's great, then you're going to
need to start to look seriously at adopting AI and
adopting AI as aggressively as you're trying to hit these
goals because there isn't any other technology, like we would
have already known about it, it would be implemented already.
(19:13):
If there was some technology, it was going to give
you double digit percentage decreases in natural gas utilization or
in you know, coal utilization, like, only AI can.
Speaker 3 (19:23):
Get you there.
Speaker 2 (19:23):
So unless you're horribly inefficient to begin with, but let's
hope that everyone is starting from at least a place
of we're trying our best.
Speaker 4 (19:33):
Well and so you know, what's really interesting you say
that because one of the questions I get is is, well,
so you know Rovis's service is fourteen different industries, so
everything from oil and gas to life science, to consumer
package goods, food and beverage, you name it. One of
the questions I get is is what industries have adopted
AI first, And it's not what you would think.
Speaker 3 (19:54):
The industries that are.
Speaker 4 (19:55):
Adopting AI first in a big way. Turns out there
the commodity industries. Glass, I've got more steel projects go
AI projects going in steel than you would imagine. And
to your point, those are processes that have been around
for like one hundred years and haven't changed that much. Right,
These are well understood, well optimized processes and they are
the one. But it's also very tight margins. It's commodity pricing.
(20:18):
So they are the ones that know if they can
squeeze a couple more percentages out of their existing equipment,
that that can be the edge over their competitors. The
industries that have not adopted it widely is life science,
I think because of the risk aversion, plus they always
have to worry about FDA validation and things like that,
so there's a there's a there's a lot more risk there,
(20:40):
and so they have not adopted it as as much
as you would think.
Speaker 3 (20:43):
And as quickly as you would think.
Speaker 2 (20:45):
Yeah, that's interesting. So when we're talking about some of
these adoptions and readiness to adopt, you mentioned that, you know,
these commodities companies, these these raw materials companies, they're they're ready,
they're chomping at the bit, They're they're interested. One of
I mean that's a critical component is being ready. You're
not going to be able to make a change if
your if your team, if your organization isn't ready for it.
(21:09):
What are some of the common pitfalls that manufacturers encounter
regarding like data quality. I know that's a huge one
for us when we're working with our clients, is the
quality of the data that they're able to provide. I'm
sure as an integrator you also face that where they
think they have good data and then they hand it
to you and you're like, way, not, it's it's okay.
(21:32):
So the data quality, things like their their infrastructure, like
what do they have to be able to work with
once they've they've got some information? And what kind of
practical steps could companies who are interested start making in
order to increase their maturity and be ready for those
next AI projects.
Speaker 4 (21:51):
Yeah exactly, And so I mean there's a lot of
pitfall pitfalls and tips and tricks and things like that
that we could talk about. But let me let me
boil down, because you already mentioned a couple let me
add to that list. So I would say the number
one thing is data infrastructure at the plant floor level.
So this is capturing the data as it happens in
(22:13):
real time. For those of you who may not be
as familiar your listeners with manufacturing, we use these what
are called time series databases that we call them historians,
and we capture that process data, the very fast moving data.
We captured it in real time and we store it
in these optimized databases that are made to store this
type of data. We want to get more historians in.
(22:35):
We want to get more data capture as close to
the sources of data as possible, So that's step one.
Step two then is that we want to consolidate that
data into the cloud. So we want to marry that data,
both the process data, but then we want to marry
it to the transactional and relational data that's both on
the plant floor but also might exist in IT systems
(22:58):
like EUERP systems.
Speaker 3 (22:59):
We need to.
Speaker 4 (22:59):
Marry a really important factor that a lot of people
underestimate is mirroring up that quality data, right, So it's
not enough to know how the thing was made if
you were making junk the whole time, So you need
to know what was the final quality of that batch.
There's some tricks that we can do to marry that
data together so that you can get a single data set.
But ultimately, and even the data quality things you were
(23:20):
talking about, those are important, but a lot of times
we've got we do real data science too, so the
data science algorithms a lot of times can tell you
when you're having data quality issues, so you can go
back and kind of fix those after the fact. But ultimately,
we're trying to build consolidated, clean, concise, consolidated data sets
that we can then leverage to build mL models. But
(23:42):
the final step or the final tip here I would
say around readiness is to start with use cases and
work backwards to the technology and the data that's required.
Most customers get that backwards, and I understand because you've
got a lot of vendors in the space that are
hammering on them technology technology, so they get technology in
their head and then they try to work backwards to
(24:03):
the use case. You really want to start with use cases,
figure out where the value is, and then work backwards
to what's the data I'm going to need to make
that a reality, and what is the technology. There's no
limit to the amount of money in time you could
spend on data and technology. The limit is what value
you can get out of it. And that's what the
use cases are going.
Speaker 3 (24:22):
To tell you. All right.
Speaker 2 (24:24):
So speaking of value and return on investment, that's critical
obviously all the time. But what other measures of success
are there with AI? And you know, is there something
else they should be looking at to measure success of
a project.
Speaker 4 (24:39):
One hundred percent? And I'm so glad you asked that
because I get into these meetings where customers get so
laser focused on the dollar ROI and it's like, yes, okay,
I understand that, and we can get there with AI projects.
We can pick the right projects that have the dollar
dollar ROI. But there are so many other things. The
first one I would say is quality. I'd say three.
(25:00):
The first one is quality of life for those folks
who actually work in the plant. So if we can
make their life better, safer, if we can put them
in positions where they're adding more value to the organization
if we can just make their life less repetitive and
their job, you know, and make them excited to come
to work.
Speaker 2 (25:19):
Like.
Speaker 4 (25:19):
Those are all things. I mean, you're not those aren't
going to be captured in dollars, but those are the
quality of life things that AI can bring. I would
say the second one is agility to face shifting regulations,
Agility to handle disruptions and supply chain, agility to respond
to customer demands. I've got a lot of customers that say,
you know, five ten years ago, we were the company
(25:41):
that like eighty percent of our output was one product,
one skew. We just we just kind of burn and
churn on this one thing. And now suddenly we've got
this huge mix of products. We've got a service because
that's where the market has gone. You know, it's more
custom it's more focused. So that ability to to to
handle those things and face those types of changes more
(26:03):
more quickly. And the final one, and the one that
I don't think manufacturers think enough about, is the ability
to attract a workforce who has grown up with technology.
So these these younger generation, they my kids were born
with an iPad in their hand, good, bad, or indifferent,
like we just were Like they've been using an iPad
since they could, you know, crawl and whatever, so so
(26:26):
they are used to that and that's what they're looking
for in their job that they're going to spend eight
plus hours a day at so so manufacturers need to
spend more time thinking about that ability to attract a workforce.
You know, it's one of those things that Silicon Valley startups,
those tech companies, they think a lot about how do
I attract and keep a good workforce and manufacturers. You know,
(26:49):
so an AI project even you know, maybe the dollar
amount ROI you know, is closed, but maybe not as compelling.
But boy, if I'm a company that looks like I
invest in AI, and I can use that as a
talking point, and I can get out to career fairs
and say, hey, we've got five AI projects going right now,
that's a company that's going to attract those workers again
that are in high demand because they don't have enough
(27:11):
of these people.
Speaker 2 (27:12):
So absolutely, because it goes right back to your first
point that the you know, we don't have the longevity
in these positions anymore, and it good better otherwise we
have to assume that that's because of some sort of churn.
And if that's the case, then you don't only need
to attract the right people, you need to keep the
(27:32):
right people. And so if that's going to elevate their
status in the eyes of today's workforce, that's huge. And
that's not necessarily something you can measure with the number.
I'm sure there's a number somewhere, a metric, a KPI
that they're going to come up with. They're going to
say that this is our you know, net quality score
or whatever for recruiting or for you know, being a
(27:55):
techie company. There's lots of awards, there's lots of distinctions
that they can utilize to show that, but it's it's
not a dollar amount that's going to go on on
their annual report. So, looking towards that next decade, the
next five to ten years, how do you see it
evolving in manufacturing? Is there a trend or a particular
(28:16):
technology within the AI scope that that manufacturers should be
watching out for to stay ahead one hundred percent?
Speaker 3 (28:23):
So obviously generative AI.
Speaker 4 (28:27):
So for those for your listeners, genera of AI is
an umbrella term that includes CHET, GPT, all the large
language models or llms as well as the.
Speaker 3 (28:35):
Image and video generation sure generative AI.
Speaker 4 (28:39):
We understand the impact it's having, and it's it's new
and it's exciting. And again, when most people say AI,
they think that's what they think.
Speaker 3 (28:46):
Of generative AI.
Speaker 4 (28:48):
Right, there are other types of AI, and one in
particular that I'd like to highlight briefly is what's called
autonomous AI. And autonomous AI is AI that's built on now,
it's almost a decade old a mL algorithm called deep
reinforcement learning. And what makes autonomous AI unique is its
ability to it's able to build long term strategy, so
(29:12):
it's able to make human like decisions. If they started
out by using it to play games, it beat all
of our best grand masters at the game of go,
at the game of chess, StarCraft. But then you know,
it's matured a lot since then, and so we're using
and I've got customers, real customers that are running on
the plant floor right now that it's they're using an
in production. And again, it's the perfect technology to because
(29:34):
it can learn like a human can. It's the perfect
technology to augment the capability of younger employees, newer employees
and through a process that we use called machine teaching.
We're actually able to capture that expertise from those subject
matter experts and build it into the autonomous AI.
Speaker 3 (29:53):
So it uses that as.
Speaker 4 (29:54):
Its basis, and then when we train it, it learns
and gets to an expert level at whatever part of
the problem us we're pointing it out. So I think
a lot of manufacturers when I talk to them, they
tend to call me in because they want to talk AI,
and they think we're going to talk about generative AI,
and immediately I pivot and I talk about autonomous AI
because the reality of it is is that the types
(30:14):
of problems that we're solving on the plant floor tend
to be more operational problems, and that's what autonomous AI
is good at addressing. JENITTI of AI is great at
knowledge type problems, but those are kind of in a
different part of the manufacturing facility. We need to solve
operational problems. So I would encourage for those manufacturers who
maybe haven't looked into autonomous AI, please do look into that.
(30:35):
It is we feel like an inflection point. We do
feel like autonomous AI and the adoption of it is
going to be one of those things that's going to
make a c change in you know, a big step
change in how our manufacturers work. So we're very bullish
on it. And then of course generative AI. I mean,
there's no question that it's going to have a huge impact,
(30:56):
and we should continue, my customers should continue to look
into generative and keep it on the radar. But there's
some very big limitations.
Speaker 3 (31:03):
With generative AI. It cannot reason.
Speaker 4 (31:06):
In fact, there was a study that came out from
Apple late last year that finally proved once and for
all that jenerat of AI is not capable of reasoning.
Speaker 3 (31:13):
It will never be a tot.
Speaker 4 (31:14):
Them say, I can reason, it can understand causal connections,
but generative I won't be able to. And the other
thing is hallucinations, which they're they're working to try to minimize.
But we just saw on Saturday, the Chicago Sun Times
ran an article, I don't know if you saw it
with the summer reading list for twenty twenty five, and
over half of the books on that list were don't exist.
(31:35):
Don't exist because the author had used AI to generate
the thing, and no one from editors on now and
no one had bothered a double check that these were
actual books, so hallucinations, I mean, and that's you know,
it's funny when it does those types of things in
those kinds of context or generates crazy marketing lurbs. But
on the plant floor, I.
Speaker 3 (31:53):
Mean, you could kill somebody.
Speaker 4 (31:54):
So it's just not appropriate to use generative AI yet
on the plant floor. And you know so, but keep
an eye on I mean, it's obviously it's going to
get there, but it's still a relatively new technology, all right.
Speaker 2 (32:06):
So it's clear that this conversation is just beginning, and
we look forward to seeing how these strategies continue to
evolve in the manufacturing space. Thanks so much, Brian for
joining us and sharing your insight today.
Speaker 3 (32:21):
Yeah, thank you Ellen.
Speaker 2 (32:22):
To our listeners, if you found today's episode valuable, be
sure to follow Speaking of Supply Chain wherever you get
your podcasts and share it with someone who's shaping the
future of operations. Until next time, keep innovating and keep automating.
Speaker 1 (32:39):
You've been listening to Speaking of Supply Chain a meboch podcast,
Keep connected with us by subscribing to the show in
your favorite podcast player. If you like what you've heard,
please rate the show that helps us to keep delivering
the latest in supply chain information. Thanks for listening. Until
next time.
You're listening to Speaking of Supply Chain, a meeboch podcast.
This is a show for logistics professionals looking to learn
more about the latest innovations in supply chain. Each episode
will feature a conversation on topics such as mitigating supply
chain disruption and reducing risk, current automation trends, sustainability initiatives,
and more. Let's dive right in.
Speaker 2 (00:30):
Hello and welcome to Speaking of Supply Chain, where we
explore trends, current events, and innovations impacting the logistics and
supply chain industries. I'm your host, Ellen Wood. AI is
rapidly transforming manufacturing by enhancing efficiency, reducing costs, and addressing
labor shortages. Recent stats indicate over eighty percent of businesses
(00:50):
have adopted AI to some extent, with thirty five percent
implementing it across multiple departments. But one area that's still
working to catch up is in dust operations. From predictive
maintenance to automating repetitive tasks, the industrial manufacturing sector is
rapidly finding ways to incorporate these advanced capabilities. Joining me
(01:11):
today to dig into some of the details is Brian boy,
director of Industrial AI at Rovisus, welcome Brian.
Speaker 3 (01:18):
Hi, thanks for having me.
Speaker 2 (01:20):
Great to have you. So before we get into our
discussion today about AI, let's take a moment just to
get to know each other a little bit better. This
is a feature that a lot of our listeners are
probably not used to. It used to only be available
in our video content, but now that video is available
on Spotify, we're putting it into the full program. So
(01:43):
our icebreaker question for today, I've got one for you,
and then you've got a different one for me. So
mine for you. What's a weird fact about you?
Speaker 4 (01:53):
So I would say the I live a pretty normal
suburban life. But I would say the one weird fact
is that my wife and I are kind of roller
coaster freaks, and so we both have each ridden I
would say at least over one hundred rides on roller coasters,
probably one hundred and fifty at this.
Speaker 3 (02:11):
Point when we met.
Speaker 4 (02:13):
We met at eighteen in college and both kind of
bonded over roller coasters. We live in Ohio and so
we've got Cedar Point is up the street from us,
so that's you know, that's kind of why. And then
when we raised our kids, we kind of raised them
to be roller coaster freaks, so they ride with us now,
so all five of us jump on these coasters.
Speaker 3 (02:32):
But yeah, we love them. We love them big and
fast and scary.
Speaker 2 (02:36):
So my kids also love roller coasters, and my husband
is a roller coaster freak. We've been to Cedar Point
multiple times. There's one roller coaster there. I forget which
one it is. It's the ginormous one that's so tall
that you can see Canada on a clear day.
Speaker 3 (02:50):
Yeah, that's money and force.
Speaker 2 (02:52):
The millennium force. So I remember, I love riding that
roller coaster. But I read something recently about it that
was really interesting, and that is that it is one
of the few roller coasters in the world that can
help break up a person's kidney stones. Totally random. If
you have kidney stones, you can ride that roller coaster
and it will help break them up so that you
(03:13):
can pass them.
Speaker 3 (03:14):
I don't know.
Speaker 4 (03:16):
I don't know how accurate that is, but uh, I
guess as I age, that's good to know.
Speaker 3 (03:20):
If it's true.
Speaker 2 (03:22):
That speaks to the force of that roller coaster.
Speaker 4 (03:24):
It is the first hill. It's wild, it's wild.
Speaker 2 (03:28):
Yeah, yeah, yeah. I was talking about roller coasters just
this past weekend with my son. We were driving through
Louisville and looking over and seeing when Kentucky Kingdom was
going to open, because it's just off I sixty five
in Louisville. You can see it right by the fairgrounds.
And he's he's very excited. He's like, when is it
going to open? Can we go find roller coasters?
Speaker 3 (03:49):
Yep?
Speaker 4 (03:50):
Yeah, I can't wait. We got our season passes. I
can't wait for see your point open.
Speaker 3 (03:54):
We'll be there.
Speaker 2 (03:55):
So all right, that sounds like a lot of fun
for the summer. So you had a different question for
me to get to know me a little bit better
for our listeners.
Speaker 3 (04:03):
And what was that I did? So, Ellen, I like
to cook a lot.
Speaker 4 (04:07):
I do all the cooking in our house, so I'm
always looking for new ideas, new inspiration.
Speaker 3 (04:11):
What's your favorite dish to cook?
Speaker 2 (04:13):
Oh? My favorite or the one I cook the most often,
because that's a different thing all together.
Speaker 3 (04:19):
I would say, what's your Yeah, what's your favorite?
Speaker 2 (04:22):
My favorite? So I love making soup, and yeah, I
make soup all the time. And I've I've worked this
past winter and fall at perfecting my bone broth. Okay,
and I've been making multiple batches of it and I
make bone broth at least once a week, and I
(04:44):
freeze it. I give it away to friends. But I
just I've wanted to find like the best bone broth,
and I like beef bone broth, and I have I
have a grocery store near us. It's an international grocery
and it has a lot of different ethnic that most
grocery stores do not have. And this store has all
(05:04):
the parts that you need for making good broth that
most grocery stores aren't going to have prepackaged nicely for
someone to go consume. And one of the biggest additions
to my chicken broth is the feet, oh chicken.
Speaker 3 (05:19):
Okay, okay.
Speaker 2 (05:22):
You get so much more collagen out of chicken feet,
and it's not something you will ever find at a
regular grocery store. But if you can get your hands
on them from a butcher and add them into your
your bones and your your vegetables and your herbals that
you put in there, that the chicken feet make all
the difference in a very very collagen rich broth, which
(05:43):
is what you want.
Speaker 3 (05:43):
Yeah, well, that's a good tip. I like that.
Speaker 2 (05:46):
It makes my kids laugh every time I make it
because there's these little chicken feet stickn open.
Speaker 3 (05:56):
Well, you know, as long as it tastes good in the.
Speaker 2 (05:58):
End, exactly, it hasn't been them from eating it. So
all right, Well, let's get on to our topic today,
which is not chicken feed or roller coasters. It is
AI in industrial manufacturing. And this is something that you
have a passion for, So tell us a little bit
about it.
Speaker 4 (06:17):
Yeah, So, well, I mean to get there, I kind
of have to start back at my journey here, my
career at Robusis. So I work for a company. So
we're a system integrator and we target manufacturing and industrial customers.
So I started here actually right out of college. I
co opt here and then I started full time. So
I've actually spent twenty five years here at Robusis has
(06:38):
been a health professional career, which you don't see as
often any more, folks spending kind of their whole career
at a single company. And I came in with a
software background, and so at the beginning of my career,
we were writing a lot of custom software and things
like that for customers. Now, things that progress, there's a
lot of commercial off the shelf software that you can buy,
(06:59):
so so in some ways we've kind of advanced and
now so we're focused on AI for those same.
Speaker 3 (07:05):
Industrial and manufacturing customers.
Speaker 4 (07:07):
I'm very passionate about it because I you know, I've
spent I really came into this industry knowing nothing about manufacturing.
Speaker 3 (07:13):
When I started twenty five years ago, I.
Speaker 4 (07:15):
Knew I liked the show How It's Made, so that
was kind of cool, like to seeing how stuff was made.
And so now I've spent the last two and a
half decades working in this industry and really have fallen
in love with manufacturing. I do feel like manufacturing is
the biggest lever that we can pull as a civilization
to create the most prosperity for the most people. I
(07:37):
think it's that important. I think it is the wealth
generation mechanism for for civilization, for societies, certainly for economies,
and so I'm.
Speaker 3 (07:47):
Very passionate about it. I'm passionate about what we do.
Speaker 4 (07:49):
And manufacturers in general are not known for being real
cutting edge right there for a lot of good reasons.
They're very risk averse, and so they're typically five to
seven years behind more of the IT world and in
the rest of the world other industries, and so it's
been you know, a longer process to get them to
(08:12):
adopt AI. But all of I mean almost all of
my customers now are taking some kind of steps towards that.
Even if they haven't gone full born AI yet, they've
got a data science team where they're at least starting
to build out some of the infrastructure, do some of
the infrastructure upgrades that are required to get to an
AI future?
Speaker 2 (08:28):
All right, So once that initial AI project gets off
the ground, how can organizations effectively scale that up and
make it into those bigger projects that don't necessarily overwhelm
the team As you said, they're they're risk averse. They're not.
This isn't their wheelhouse, this isn't where they look for
(08:50):
the next thing, and this is this is very new.
So how do we do it without overwhelming those teams
or risking the quality of that implementation and try to
do something We certainly don't want to do something and
fail because then you know, it's it's going to make
the team risk averse to trying something new again.
Speaker 3 (09:08):
Right, Yeah for sure?
Speaker 4 (09:10):
And I know it sounds self serving, but the answer
really is to hire a system integrator. So that's that's
really our role is to in cases where the customer
doesn't have any internal capability like that, any internal team.
We're there to bring in the experts and to be
the horsepower behind those types of projects. But when we
also have plenty of customers that have existing you know,
(09:33):
data science or AI teams and then we come in
to scale their capabilities out. So you know, it's like, look,
we want to you know, we've heard this where we've
got some models built and we think we were onto
something here, but we want to roll this out to
five sites. So we've got a three year plan to
roll us out to five sites. I'm like, wait, I
can scale horizontally, I can bring in bodies and we
(09:55):
can do we can roll it out to five sites
in parallel. And so that suddenly is is an eye
opener there when they realize we.
Speaker 3 (10:04):
Could shrink this schedule a heck.
Speaker 4 (10:05):
Of a lot more So, like I said, I know
it's sound self serving, but that really is the role
that system integrators play in these types of things is
to augment that capability.
Speaker 3 (10:14):
I will say.
Speaker 4 (10:14):
Longer term though, one of our key tenants at robsis
is that we believe that customers should own the solution.
Speaker 3 (10:22):
When we're done. So we spend a lot of time
on training.
Speaker 4 (10:26):
You know, we tell customers you're allowed to look over
our shoulder the entire project if you'd like. We want
to make sure that you feel comfortable owning this system
when when we're done, because I mean, it is your
system ultimately, and that's that you know. So we you know,
we know that they're going to call us when they
have the next big project or whatever. But we don't
want to lock them in into a situation where, you know,
every time they have to retrain the model or add
(10:49):
a little you know, add some tags to it or something,
we don't have to be able to force to call
us for that kind of thing.
Speaker 2 (10:56):
Sure, so as it becomes more embedded in these workforces,
as you're handing off the capabilities and the tools to
do it, what new skills should companies be proactively you know,
teaching their employees and getting people to a comfort level
with this type of technology, and how can they best
(11:17):
approach the training for those things? Because I know we've
had a number of trainings in our organization on AI.
We had a workshop just yesterday trying to you know,
workshop and get some new ideas, some fresh ideas that
aren't just what's your chat GPT prompt? I mean, that's
that's the one I use the most. But how do
(11:38):
we train this next generation of workers to function with
the AI and not fear it.
Speaker 4 (11:47):
Yeah, so there's a couple of different answers there, because
it really does depend on what your role is in
the organization. So if you're somebody who actually feasibly you're
an analyst or an engineer, and you feasibly would be
maybe interfacing to these models or maybe even building these models,
you're going to need some kind of data science background. Now,
(12:07):
the good news is that there are free resources. There
are in resources available, so there's lots of ways to
get some of that data science background. I will say though,
that that doesn't seem to be as much of a
problem for our customers where they're seeing the bigger problem
is in the loss of expertise. So we are seeing
(12:28):
some people are calling it a silver tsunami, but we
are seeing a loss of primarily baby boomers post COVID
who are leaving the manufacturing workforce, and what we are
not seeing is a corresponding increase in younger folks entering
the manufacturing workforce. So just to give you some SPAT stats.
This came from LNS Research in twenty nineteen. The average
(12:51):
tenure in the this is for US. The average tenure
for the manufacturing workforce average years at a company is
twenty years. The average tenure, So that was twenty nineteen
and twenty twenty three, that average tenure fell to three years. Yeah,
the average time and position fell from seven years to
(13:12):
nine months in twenty twenty three, in just four years.
It's a significant issue that we're seeing. It is the
number one issue that customers bring up when we sit
down with them nowadays. It's a continuing issue. And so
there's two aspects of that then that we typically talk about.
One of those things is how can we build AI
(13:33):
systems that can augment the capabilities of a novice person
someone who hasn't been at the place very long. They
also have very high turnover in these positions, so oftentimes
you're talking about operators on the line. You know, you
had a ten to fifteen year person and the person
replacing them maybe it has six months to a year
(13:53):
in that position. Because the high turnover, they're constantly starting
over again. So how can we build AI systems that'll
supplement the capabilities of these no office operators so that
we're not just running these lines suboptimally forever. And the
other aspect of it is capturing that expertise from their
subject matter experts before they leave. So I was just
(14:16):
at a conference. I was speaking at a conference this
week and that was one of the key things they said.
Somebody raised their hands said, you know, we've got We've
got some of those people that you're talking about that
have been here for fifteen twenty years, but they are
on their way out. I mean within the next couple
of years they're playing and retiring. What can we do
to capture that expertise in AI? And the good news
is is that there are mechanisms to mechanisms to do that,
(14:38):
but the time to do that is now, before you
lose that expertise out the door.
Speaker 2 (14:42):
Yeah, that really has to be a proactive program to
get that information while there's still the opportunity to do so.
And it's great that they're AI tools to do that
as well. But one of the things you were talking
about earlier was, you know, driving the efficiency and the
productivity within that manufacturing facility. How can they also leverage
(15:08):
that AI to advance sustainability goals or something like energy
management or waste reduction. You know, there are a lot
of other things besides, just like you said, the manufacturing
process itself not that excited about changing the way they
do it. They're five to seven years behind. So there
(15:31):
are other ways that they can use this technology to
improve operations. That isn't just the manufacturing process, but more
about the facility itself and the processes of or I
guess the byproducts of the processes. There's so much involved
in manufacturing and AI is going to touch all of it.
Speaker 4 (15:50):
Yeah, yeah, And so it's actually it's so interesting to
kind of generationally to see because especially when you talk
to millennials, when you talk to gen z's sustainabilities top
of mind for them, and many of them are now
in leadership positions at these companies, and so that's good.
Like so, so we want to be cognizantve sustainability. There's
(16:11):
kind of two aspects there. One is that we need
to start to define AI projects as sustainability projects, because
you mentioned a whole bunch of things, and every single
one of those things. If I'm increasing the efficiency of
the line, I am implicitly I am reducing energy. I
am I am increasing throughput, which means that I'm reducing
waste all of those things. I mean, the amount of
(16:34):
waste in manufacturing is mind bibling when you see how
much gets thrown out, how much draft gets thrown out, uh,
and then the amount of energy that's used to run
these facilities. So if I'm doing an AI project and
I'm squeezing you know, singlar double digit percentage improvements out
(16:55):
of those lines, I am inevitably reducing the amount of energy.
Speaker 3 (16:59):
You know.
Speaker 4 (17:00):
We did an AI project for a glass manufacturer, and
it's so funny too, because depending on who you talk
to at the company, you tend to get a different answer.
So we did a glass project and we were building
an AI optimizer to more efficiently cut glass. Think like
window panes out of a big sheet of glass. Obviously,
everything else that's not window is scrap. It gets broken
(17:20):
down and it gets remelted. Now, we were talking to
a couple of the operators and they go, well, we
don't have scrap because all of that gets that gets remelted, right, So,
so effectively our scrap level is zero. Then later on
in the day, you have a meeting with one of
the finance people and they're like, no, all of that
is scrap. We track all of that, even if they
don't know, we track all of that because all of that,
(17:42):
even the melting down. You think about all of the
energy that went into producing that pane of glass the
first time, the first time. Now you're gonna break it down,
you're gonna smash it, and you're gonna remelt it and
pour all of that energy and raw material back into it,
all the labor that it took. Absolutely so all of
that is wasted energy. So we were made able to
make massive Their optimization engine that they were using to
(18:04):
lay out these glass panes was very suboptimal. We were
able to squeeze so many more panes of glass out
of out of these sheets of glass and reduce their
scrap by double digit percentage. That's a huge impact on
energy reduction. Was that classified as a sustainability project by
corporate It was not.
Speaker 3 (18:20):
It should have been.
Speaker 4 (18:21):
So that's what I'm saying, we need to start defining
these AI projects as sustainability projects.
Speaker 3 (18:26):
But the other aspect of it is.
Speaker 4 (18:28):
That I think is in some ways even more important
when we look at sustainability goals, is many of these
companies have very aggressive sustainability goals. Great, we want to
be carbon neutral bi X here. Fine, here's the problem
with that is there is no other technology other than
AI that is like in the wings waiting to reduce.
Speaker 3 (18:48):
Their greenhouse impact or their carbon like.
Speaker 4 (18:51):
There isn't any other technologies other than AI. So if
you want to get hit these aggressive state sustainability goals,
which again I encourag that's great, then you're going to
need to start to look seriously at adopting AI and
adopting AI as aggressively as you're trying to hit these
goals because there isn't any other technology, like we would
have already known about it, it would be implemented already.
(19:13):
If there was some technology, it was going to give
you double digit percentage decreases in natural gas utilization or
in you know, coal utilization, like, only AI can.
Speaker 3 (19:23):
Get you there.
Speaker 2 (19:23):
So unless you're horribly inefficient to begin with, but let's
hope that everyone is starting from at least a place
of we're trying our best.
Speaker 4 (19:33):
Well and so you know, what's really interesting you say
that because one of the questions I get is is, well,
so you know Rovis's service is fourteen different industries, so
everything from oil and gas to life science, to consumer
package goods, food and beverage, you name it. One of
the questions I get is is what industries have adopted
AI first, And it's not what you would think.
Speaker 3 (19:54):
The industries that are.
Speaker 4 (19:55):
Adopting AI first in a big way. Turns out there
the commodity industries. Glass, I've got more steel projects go
AI projects going in steel than you would imagine. And
to your point, those are processes that have been around
for like one hundred years and haven't changed that much. Right,
These are well understood, well optimized processes and they are
the one. But it's also very tight margins. It's commodity pricing.
(20:18):
So they are the ones that know if they can
squeeze a couple more percentages out of their existing equipment,
that that can be the edge over their competitors. The
industries that have not adopted it widely is life science,
I think because of the risk aversion, plus they always
have to worry about FDA validation and things like that,
so there's a there's a there's a lot more risk there,
(20:40):
and so they have not adopted it as as much
as you would think.
Speaker 3 (20:43):
And as quickly as you would think.
Speaker 2 (20:45):
Yeah, that's interesting. So when we're talking about some of
these adoptions and readiness to adopt, you mentioned that, you know,
these commodities companies, these these raw materials companies, they're they're ready,
they're chomping at the bit, They're they're interested. One of
I mean that's a critical component is being ready. You're
not going to be able to make a change if
your if your team, if your organization isn't ready for it.
(21:09):
What are some of the common pitfalls that manufacturers encounter
regarding like data quality. I know that's a huge one
for us when we're working with our clients, is the
quality of the data that they're able to provide. I'm
sure as an integrator you also face that where they
think they have good data and then they hand it
to you and you're like, way, not, it's it's okay.
(21:32):
So the data quality, things like their their infrastructure, like
what do they have to be able to work with
once they've they've got some information? And what kind of
practical steps could companies who are interested start making in
order to increase their maturity and be ready for those
next AI projects.
Speaker 4 (21:51):
Yeah exactly, And so I mean there's a lot of
pitfall pitfalls and tips and tricks and things like that
that we could talk about. But let me let me
boil down, because you already mentioned a couple let me
add to that list. So I would say the number
one thing is data infrastructure at the plant floor level.
So this is capturing the data as it happens in
(22:13):
real time. For those of you who may not be
as familiar your listeners with manufacturing, we use these what
are called time series databases that we call them historians,
and we capture that process data, the very fast moving data.
We captured it in real time and we store it
in these optimized databases that are made to store this
type of data. We want to get more historians in.
(22:35):
We want to get more data capture as close to
the sources of data as possible, So that's step one.
Step two then is that we want to consolidate that
data into the cloud. So we want to marry that data,
both the process data, but then we want to marry
it to the transactional and relational data that's both on
the plant floor but also might exist in IT systems
(22:58):
like EUERP systems.
Speaker 3 (22:59):
We need to.
Speaker 4 (22:59):
Marry a really important factor that a lot of people
underestimate is mirroring up that quality data, right, So it's
not enough to know how the thing was made if
you were making junk the whole time, So you need
to know what was the final quality of that batch.
There's some tricks that we can do to marry that
data together so that you can get a single data set.
But ultimately, and even the data quality things you were
(23:20):
talking about, those are important, but a lot of times
we've got we do real data science too, so the
data science algorithms a lot of times can tell you
when you're having data quality issues, so you can go
back and kind of fix those after the fact. But ultimately,
we're trying to build consolidated, clean, concise, consolidated data sets
that we can then leverage to build mL models. But
(23:42):
the final step or the final tip here I would
say around readiness is to start with use cases and
work backwards to the technology and the data that's required.
Most customers get that backwards, and I understand because you've
got a lot of vendors in the space that are
hammering on them technology technology, so they get technology in
their head and then they try to work backwards to
(24:03):
the use case. You really want to start with use cases,
figure out where the value is, and then work backwards
to what's the data I'm going to need to make
that a reality, and what is the technology. There's no
limit to the amount of money in time you could
spend on data and technology. The limit is what value
you can get out of it. And that's what the
use cases are going.
Speaker 3 (24:22):
To tell you. All right.
Speaker 2 (24:24):
So speaking of value and return on investment, that's critical
obviously all the time. But what other measures of success
are there with AI? And you know, is there something
else they should be looking at to measure success of
a project.
Speaker 4 (24:39):
One hundred percent? And I'm so glad you asked that
because I get into these meetings where customers get so
laser focused on the dollar ROI and it's like, yes, okay,
I understand that, and we can get there with AI projects.
We can pick the right projects that have the dollar
dollar ROI. But there are so many other things. The
first one I would say is quality. I'd say three.
(25:00):
The first one is quality of life for those folks
who actually work in the plant. So if we can
make their life better, safer, if we can put them
in positions where they're adding more value to the organization
if we can just make their life less repetitive and
their job, you know, and make them excited to come
to work.
Speaker 2 (25:19):
Like.
Speaker 4 (25:19):
Those are all things. I mean, you're not those aren't
going to be captured in dollars, but those are the
quality of life things that AI can bring. I would
say the second one is agility to face shifting regulations,
Agility to handle disruptions and supply chain, agility to respond
to customer demands. I've got a lot of customers that say,
you know, five ten years ago, we were the company
(25:41):
that like eighty percent of our output was one product,
one skew. We just we just kind of burn and
churn on this one thing. And now suddenly we've got
this huge mix of products. We've got a service because
that's where the market has gone. You know, it's more
custom it's more focused. So that ability to to to
handle those things and face those types of changes more
(26:03):
more quickly. And the final one, and the one that
I don't think manufacturers think enough about, is the ability
to attract a workforce who has grown up with technology.
So these these younger generation, they my kids were born
with an iPad in their hand, good, bad, or indifferent,
like we just were Like they've been using an iPad
since they could, you know, crawl and whatever, so so
(26:26):
they are used to that and that's what they're looking
for in their job that they're going to spend eight
plus hours a day at so so manufacturers need to
spend more time thinking about that ability to attract a workforce.
You know, it's one of those things that Silicon Valley startups,
those tech companies, they think a lot about how do
I attract and keep a good workforce and manufacturers. You know,
(26:49):
so an AI project even you know, maybe the dollar
amount ROI you know, is closed, but maybe not as compelling.
But boy, if I'm a company that looks like I
invest in AI, and I can use that as a
talking point, and I can get out to career fairs
and say, hey, we've got five AI projects going right now,
that's a company that's going to attract those workers again
that are in high demand because they don't have enough
(27:11):
of these people.
Speaker 2 (27:12):
So absolutely, because it goes right back to your first
point that the you know, we don't have the longevity
in these positions anymore, and it good better otherwise we
have to assume that that's because of some sort of churn.
And if that's the case, then you don't only need
to attract the right people, you need to keep the
(27:32):
right people. And so if that's going to elevate their
status in the eyes of today's workforce, that's huge. And
that's not necessarily something you can measure with the number.
I'm sure there's a number somewhere, a metric, a KPI
that they're going to come up with. They're going to
say that this is our you know, net quality score
or whatever for recruiting or for you know, being a
(27:55):
techie company. There's lots of awards, there's lots of distinctions
that they can utilize to show that, but it's it's
not a dollar amount that's going to go on on
their annual report. So, looking towards that next decade, the
next five to ten years, how do you see it
evolving in manufacturing? Is there a trend or a particular
(28:16):
technology within the AI scope that that manufacturers should be
watching out for to stay ahead one hundred percent?
Speaker 3 (28:23):
So obviously generative AI.
Speaker 4 (28:27):
So for those for your listeners, genera of AI is
an umbrella term that includes CHET, GPT, all the large
language models or llms as well as the.
Speaker 3 (28:35):
Image and video generation sure generative AI.
Speaker 4 (28:39):
We understand the impact it's having, and it's it's new
and it's exciting. And again, when most people say AI,
they think that's what they think.
Speaker 3 (28:46):
Of generative AI.
Speaker 4 (28:48):
Right, there are other types of AI, and one in
particular that I'd like to highlight briefly is what's called
autonomous AI. And autonomous AI is AI that's built on now,
it's almost a decade old a mL algorithm called deep
reinforcement learning. And what makes autonomous AI unique is its
ability to it's able to build long term strategy, so
(29:12):
it's able to make human like decisions. If they started
out by using it to play games, it beat all
of our best grand masters at the game of go,
at the game of chess, StarCraft. But then you know,
it's matured a lot since then, and so we're using
and I've got customers, real customers that are running on
the plant floor right now that it's they're using an
in production. And again, it's the perfect technology to because
(29:34):
it can learn like a human can. It's the perfect
technology to augment the capability of younger employees, newer employees
and through a process that we use called machine teaching.
We're actually able to capture that expertise from those subject
matter experts and build it into the autonomous AI.
Speaker 3 (29:53):
So it uses that as.
Speaker 4 (29:54):
Its basis, and then when we train it, it learns
and gets to an expert level at whatever part of
the problem us we're pointing it out. So I think
a lot of manufacturers when I talk to them, they
tend to call me in because they want to talk AI,
and they think we're going to talk about generative AI,
and immediately I pivot and I talk about autonomous AI
because the reality of it is is that the types
(30:14):
of problems that we're solving on the plant floor tend
to be more operational problems, and that's what autonomous AI
is good at addressing. JENITTI of AI is great at
knowledge type problems, but those are kind of in a
different part of the manufacturing facility. We need to solve
operational problems. So I would encourage for those manufacturers who
maybe haven't looked into autonomous AI, please do look into that.
(30:35):
It is we feel like an inflection point. We do
feel like autonomous AI and the adoption of it is
going to be one of those things that's going to
make a c change in you know, a big step
change in how our manufacturers work. So we're very bullish
on it. And then of course generative AI. I mean,
there's no question that it's going to have a huge impact,
(30:56):
and we should continue, my customers should continue to look
into generative and keep it on the radar. But there's
some very big limitations.
Speaker 3 (31:03):
With generative AI. It cannot reason.
Speaker 4 (31:06):
In fact, there was a study that came out from
Apple late last year that finally proved once and for
all that jenerat of AI is not capable of reasoning.
Speaker 3 (31:13):
It will never be a tot.
Speaker 4 (31:14):
Them say, I can reason, it can understand causal connections,
but generative I won't be able to. And the other
thing is hallucinations, which they're they're working to try to minimize.
But we just saw on Saturday, the Chicago Sun Times
ran an article, I don't know if you saw it
with the summer reading list for twenty twenty five, and
over half of the books on that list were don't exist.
(31:35):
Don't exist because the author had used AI to generate
the thing, and no one from editors on now and
no one had bothered a double check that these were
actual books, so hallucinations, I mean, and that's you know,
it's funny when it does those types of things in
those kinds of context or generates crazy marketing lurbs. But
on the plant floor, I.
Speaker 3 (31:53):
Mean, you could kill somebody.
Speaker 4 (31:54):
So it's just not appropriate to use generative AI yet
on the plant floor. And you know so, but keep
an eye on I mean, it's obviously it's going to
get there, but it's still a relatively new technology, all right.
Speaker 2 (32:06):
So it's clear that this conversation is just beginning, and
we look forward to seeing how these strategies continue to
evolve in the manufacturing space. Thanks so much, Brian for
joining us and sharing your insight today.
Speaker 3 (32:21):
Yeah, thank you Ellen.
Speaker 2 (32:22):
To our listeners, if you found today's episode valuable, be
sure to follow Speaking of Supply Chain wherever you get
your podcasts and share it with someone who's shaping the
future of operations. Until next time, keep innovating and keep automating.
Speaker 1 (32:39):
You've been listening to Speaking of Supply Chain a meboch podcast,
Keep connected with us by subscribing to the show in
your favorite podcast player. If you like what you've heard,
please rate the show that helps us to keep delivering
the latest in supply chain information. Thanks for listening. Until
next time.
Popular Podcasts
On Purpose with Jay Shetty
I’m Jay Shetty host of On Purpose the worlds #1 Mental Health podcast and I’m so grateful you found us. I started this podcast 5 years ago to invite you into conversations and workshops that are designed to help make you happier, healthier and more healed. I believe that when you (yes you) feel seen, heard and understood you’re able to deal with relationship struggles, work challenges and life’s ups and downs with more ease and grace. I interview experts, celebrities, thought leaders and athletes so that we can grow our mindset, build better habits and uncover a side of them we’ve never seen before. New episodes every Monday and Friday. Your support means the world to me and I don’t take it for granted — click the follow button and leave a review to help us spread the love with On Purpose. I can’t wait for you to listen to your first or 500th episode!
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
The Joe Rogan Experience
The official podcast of comedian Joe Rogan.