Subscribe
Share
Share
Embed
Watts in Your Data, hosted by Denis Gontcharov, explores how enterprises in energy & utilities leverage Databricks to improve operations. Listeners can expect in-depth technical discussions and interview that break down complex topics automated data quality testing, and advanced analytics into understandable segments, actionable insights, and real-world applications.
More About Me: https://gontcharov.eu
Denis: Hey, and welcome to the
What's in Your Data podcast.
I'm Deni Kof, your host, and
today I'm joined by Jim Gavigan.
Jim, welcome to the show.
Jim: Hey, thanks for having me.
Denis: So today we'll talk about
Ude, which is an interesting concept
developed by Jim and his company.
But before we get started, Jim,
for those of you that don't know,
don't know you, can you tell us a
bit more about you and your company?
Jim: Yeah, so.
started in the
Denis: Okay.
Jim: industrial business and
in manufacturing in 1995.
So in my 30th year, maybe coming even
on the 31st, I, I don't even know.
I think I started around the fall
of 95 and really what I started
doing was vibration analysis.
That was the very first thing I did.
Um, after a two year hiatus after
school, 'cause it was a really bad.
You know, time in, in 1993 when I
graduated, but got into vibration
analysis, did that for three months and
then I transitioned into control systems.
Whole funny story with that.
I won't bore everybody with that
one, but it is kind of funny.
But really, those first six years I did
everything from designing and building
control panels, going in and helping
'em get mounted, running conduit to 'em.
I didn't do that.
I helped, uh, helped pull wire in.
Hooked the wires up, programmed.
The PLC made the machine run
often had to go work on equipment
that I wasn't even familiar with.
Just had some of the equipment we sold
on it and hey, it doesn't run come help.
So there was no telling what I
walk would walk in on, right.
And so it really gave me a
huge base to kind of work from.
I got to see a lot of different processes,
a lot of different pieces of equipment,
and I, I would put my first six years
up against most people's, 10 or 15.
And just because of the variety I got
to see and what all I was asked to do.
'cause it, it was a lot.
And then I actually got out of the
engineering ranks and went into
sales, went to work for Rockwell
Automation as a sales engineer.
Um, definitely realized that it wasn't
just data dumping somebody about here's
all the great features of a ControlLogix,
now don't you wanna buy a hundred of them?
Right?
And they're like, well, I
can't use those right now.
I have a whole plant full of PLC five and.
hose running everywhere.
I'll eventually get there,
but I can't do it now.
And I'm like, what do
you mean these are great?
I'm like, yeah, I know
they're great, but I, use 'em.
And so I really kind of had to
learn like why people buy things.
Right.
And so that was a huge,
you know, lesson there.
Did a little technical consulting
role at the end of the Rockwell stent.
I um, recruited by a system integrator
that I worked with to be their sales
and business development person,
that's where I first learned about pie.
We were doing some work for a, uh,
food and beverage customer and um,
actually they were making cattle feed
and we ended up optimizing a process
that we had been fighting for quite
some time, maybe like 12, 18 months.
We optimize their batch process,
help them land a new customer.
And they built a, a whole new plant.
And they literally looked at
us and said, Hey, you guys
figured out what issues we had.
You know, our process, you're
helping us build the plant.
we had used PI to diagnose something.
We'd been fighting for 18 months and
was like, wow, this is really powerful.
And then when they came and asked us to
build a new plant with them, I said, two
companies just made millions of dollars.
This data thing, 'cause I
didn't know what to call it.
Just did that.
Right.
And so I knew there was something to it.
This was like probably 2010, 11 timeframe.
And then 2013 I got
recruited to go to O OSIsoft.
I was living in the Memphis, Tennessee
area at the time, so I covered two
states, Tennessee and Arkansas.
Had some corporate accounts and um,
ended up moving down to Florida,
took a a strategic account manager
role, and moved down to St.
Augustine, Florida, just
south of Jacksonville.
And then after three bosses in
two years, I got on one of those.
We're gonna fire you.
I mean performance improvement plans.
Denis: Right.
Jim: I was on one of those, right?
And so I left and went back to that system
integrator, started their PI practice.
Realized after about 18 months that
kind of needed to do this on my own.
I have some ideas and some
things you know, that I wanna do.
And here's the thing, I'd come
with a project and it's like,
okay, I think this is gonna take
four or five weeks to complete.
Right?
They're like, what's the schedule?
I'm like, we're creating
something we've never done.
I have no idea.
It could be four months, it could
be, it could be four or five weeks.
I don't know.
Right.
Because they're used to,
okay, we have an outage.
We have to be running again by this point.
So everything works back from that.
Here it's like we're creating
something out, nothing, and it's
not necessarily mission critical.
Right.
They had no concept of how this.
right.
And there was this
culture conflict, right?
And so I was like, I need
to do this on my own.
I got some ideas.
So I started Industrial
Insight in December of 16, so
coming up on nine years now.
And really at the time it was
how can I help people make the
most outta their pie systems?
'cause that was my big frustration
at O OSIsoft was I would.
Go in and tell people this.
These are all the things we're working on.
You know, pie Asset Framework was
coming of age, you know, pie Vision, you
know, had was starting to come of age.
We had built some other
technologies that, you know,
were useful and nobody was using.
I would go in and talk about them
and I got one of two answers.
Jim, that looks great.
We don't have time.
Jim, that looks great for somebody else.
I don't know how it applies to us.
And in both cases I'm like, I literally
see, because of my technical background,
I see the solution in my head, I,
I can see how this will help you.
I don't understand how, I
can't make you see that.
So it was really about how can I help
people make the most of that, right?
And then it has morphed into.
You know, we do all of that
still, but, you know, we're
working with other platforms.
You know, we work with, we've worked
with Canary, we've worked with, uh,
data Park, uh, we're working with Flow.
Um, let's see who else.
We're doing some advanced analytics stuff.
We just signed a, a
partnership deal with Seek.
Um, we're gonna look at Twin thread.
Um, we've been doing Simco for
multivariate analysis for a long time.
We've been doing business intelligence
work like Tableau and Power BI
for probably eight years now.
Denis: Okay.
Jim: So, know, really what
our goal is, is if you have.
floor data more than likely time
series, but 'cause we really don't do
MES, but we'll pull that MES data and
tie some things together, power bi.
But really from the time series
data standpoint, it touches
that, we're here to help you.
we're here to fill in any gaps you have.
Some customers are like, Hey, I just need
help on the administrative side, others
are like, need help building solutions.
'cause I don't have
people that can do that.
I can manage the system but I, and keep
it online and, you know, add tags and
you know, add new users, et cetera.
But I can't build a solution.
I don't have people to do that.
we just come in wherever
people are and help.
it varies with, with a lot of companies.
So.
So really we're here to
help with all of that.
There's, we're a team of five right now.
will grow, you know, early next year.
The way 26 is looking, you know,
hopefully that, you know, continues.
it's looking pretty positive right now.
Um, kind of working all the major
industries, pump and paper, oil
and gas, chemicals, food and bev
mining, little bit of pharma.
You know, we, we don't like to do
all the validation stuff, so we
kind of keep that to a minimum, but.
You know, probably the only one we really
don't mess with is water and wastewater,
and I don't really know why, you know,
it's, we've just never really been asked.
But pretty much all the others we're,
we're in somewhere another power gen.
We're in that one a little
bit with power distribution.
So, you know, reality is, it
makes us pretty versatile.
You know, we work on a
lot of different stuff.
We see a lot of different
things and you know, everybody
has kind of the same problems.
They just, their terminology is different.
So
Denis: Yeah.
Mm-hmm.
Well, fascinating.
I think, uh, me, me as well, I'm
also mostly focused on the level
two data from SCADA and so on.
And I should mention to the listener, the
way I actually got to know Jim was through
our mutual friend Lonnie Bowling from
the, uh, pie Group, and he was actually
my first official guest on this podcast.
So now he's great to have,
um, Jim here as well.
Jim: Yep.
Denis: So in that sense, I think,
Jim, for your discussion, it's obvious
you have a wealth of experience.
Time series data, and the
topic of this podcast really
was your new idea about ude.
Perhaps you can, before sketching
very shortly, what it is,
describe what the need for it is.
What is the origin of the
story, what's the problem?
Jim: Yeah, so it's, it's
kind of interesting.
We, we've gone through.
Um, know, reasonably
significant rebranding.
We have a marketing firm called
Group Five West outta Little
Rock, Arkansas that we work with.
I gotta give them a plug.
They've been great to work with.
So they, they need, they deserve
a lot of credit for this, but they
came out with a concept, you know,
they were looking for a fun and
creative way to describe things we do.
And they came up with this concept of ude.
And, you know, at first we were
all like, nah, but then we started
kind of having fun with the name.
Like, you need to change your attitude,
you know, like about whatever.
Right?
So we, we were having fun with it and
we're like, you know, it's kind of,
it's, it's kind of corny and, but it's
kind of cute, you know, and it kind of,
Denis: It is memorable.
Yeah.
Jim: yeah.
It's memorable.
Right?
And that's really what you want
is, is it's, it's memorable and
hopefully, you know, a good memory.
Right.
You know, hopefully
it's not something like.
me, like the Geico Gecko or the caveman,
even though I use that insurance
company, I hate their commercials, right?
I think they're silly,
but I remember them.
Um, didn't prevent me
from using their stuff.
Right.
Um, but the, the idea really comes
from, we see so many customers are
trying to these very sophisticated
systems on top of really crappy data.
And
Denis: Yeah.
Jim: of the structure in place.
Right.
And so I think the real, the
first real idea of that was,
okay, are you even ready for that?
Right?
And that was kind of what the
data to thing I think was all
about and kind of what they were
thinking and kind of how we took it.
And so we're kind of running with it.
Denis: Mm-hmm.
I think that's a very interesting
point, and you mentioned data.
In fact, one of my services also involves.
Just checking if the data is
even there, if it's good enough.
But the way from you described ude,
it goes even beyond that where data is
just one of the potential, uh, axis.
Jim: Yep.
Denis: If you can now, let's say,
summarize the idea, what exactly
would you check to measure ude?
Uh.
Jim: Yeah, so we're, we're actually
working on, um, we, we've kind of
had a struggle sometimes because
we do so much custom work, right?
How, how can we package up things
we in a smaller chunk where it
gets us in helping a customer out.
At maybe a lower price point, they
know exactly what they're gonna get.
We know exactly what we need to deliver.
And it's, it's just low barrier to entry.
Right?
And, and we figure out like, do we
like working with this customer?
They figure out do they
like working with us?
Are there other opportunities?
Things of that nature.
So we were actually, how this,
there was an idea came up of, and I
think this is where you actually up
with a concept of this podcast was.
Well, why don't we give someone
a UDE score, Like, what is
your score for your ude?
And so the, the thought process
of that was how do we build an
assessment of what does that look like?
Right?
And so I, I threw out a structure.
I'm gonna, so if you see me looking
off camera, I'm kind of, you
know, flipping through the notes.
Some of the things I think about
right is, you know, first is
from a network infrastructure.
Do you even have everything?
On a network where we
can go get to the data?
Or is it in a, in a way, even if
we can throw in like some kind
of gateway, can we get, get to
all the data that you care about?
Right.
That's a big one.
Um, from a data infrastructure
side, you know, do you have
all the right systems in place?
Do you have a data historian?
If you're a process customer, do you have
a limbs or a quality system in place?
Do you have MES?
Do you have an ERP?
Do you have a maintenance system?
Which often goes with the ERP system.
And what are those systems, you know,
are they a black box to you or is it
something we can actually get to culture?
Is your culture even ready?
For being data-driven, right.
Operations, management and engineering.
Like your engineering people
might be your operations.
People are like, no, this is
the way we've always done it.
Don't mess with me.
You know, management is like,
Hey, we want all the latest stuff.
Right?
The operations people
like, I don't want it.
Engineering's kinda like,
yeah, but we want this.
You know?
And like, are you even on the same page?
Right.
Is, is your culture even
Are they starving for it?
Denis: Yeah, I, if I can just
interrupt you and go deeper on that
point, because that's something
I've also seen in my experience.
You mentioned the different industries
you're active in, like paper,
oil and gas, um, manufacturing.
Do you also see a very big difference
in, let's say, data desire, if I
can call it this way, I found that
some enterprises, let's say the more
metallurgical manufacturing, don't
care as much about doing stuff with
data than, let's say wind energy.
'cause they have a clear demand for it.
Let's say for reporting, is that
also a part of your, um, scoring?
Jim: To, to a certain degree.
You know, a lot of the companies
we work with and, and it's, and
it's so funny because they all
point to the other one and say,
oh, see, they got it all together.
You know, like the pulp and paper
guys will say, yeah, but we wanna
be more like the oil and gas guys.
They got it all together and then
the oil and gas guys are like, well,
we're really freaking far behind.
We we're looking at the, the power
people, you know, kinda like what you
were saying, power people are like, oh,
we're looking at the food and bev folks.
You know, it's, it, it's so funny 'cause
it's, it's the fear of missing out, right?
Fomo.
Um, I think pretty much all
the industries are struggling.
I, I think, you know, there,
there's pockets, right?
I think, you know, you mentioned, you
know, wind energy and solar, which
you have a lot of the same thing.
It, it much more lends itself, you know,
Denis: Hmm.
Jim: to this kind of work because
you can scale it faster, you know?
But we go in a chemical plant, you
know, there's one that I was, I
was, I'm doing some training for
them and I'm working a little bit
on some of the curriculum stuff.
It is so.
That there's no other plant like
it in anywhere in the world.
And this is one of only a few places in
the world these products are even made.
So in some cases, you know,
like how do you take those
learnings and go elsewhere, right?
And so sometimes industries like
that kind of get stuck, especially if
they don't have a ton of competition
or even if they do like to me.
Pulp and paper's.
One of the worst at this, right, is,
well, this is the way we've always done it
Denis: Mm.
Mm-hmm.
Jim: and or don't have time.
Like I am literally firefighting all
day and I'm here for 12 hours a day, six
days a week, and I'm getting calls in the
middle of the night and all my day off.
What do you mean now you want
me to work on this stuff?
I don't have time.
Right.
And, and I think that's,
that's a huge barrier, right?
It sometimes it's not even the
desire, it's just they can't
imagine something different.
And
Denis: Right.
Jim: that's kind of a problem.
But, you know, but culturally, most
industries are kind of in the same boat.
You know, as I said, they all kind of
talk about a lot of the same things.
The, the language is different, you
know, obviously, you know, there's
a little nuances to their process
or their, know, their stuff, but.
They're all talking about
some of the same things.
They all have the issue of they're losing
gray hair experience and they're getting
a bunch of young people in and they're
trying to figure out how to still survive.
So all the, it's kind of all the same.
So, no, I, I, I don't think one
industry is better than the other.
Denis: Mm-hmm.
Well, it's good to hear that
they all have the desire.
So let, let's focus back on the concept
you mentioned network, the systems.
Um, the culture, I guess data is also
a very big important part of it, right?
Jim: Yeah.
Denis: quality.
Jim: you know, yeah, data quality
Denis: Mm-hmm.
Jim: a, is is a big one.
Um, cybersecurity, you know, kind of
what's their take on like semantic
models, like a UNS or, or if
you're their pie house, like asset
framework, you know, things like that.
You know, what are they
thinking about for all of that.
Um.
And I think one of the big ones too
is, is do they just have a lot of
first principles, components in place?
For instance, do they have good
dashboards in place to tell them
even how they're performing and if
they're not performing well, or even
if their individual processes aren't
performing well, how do you tell
the operator what to potentially do?
Do they even have those kind of things
in place from a first principle?
Right.
No, no magic to it, right?
It's just if this KPI starts
going up, then we know we need to
twist these two knobs down, right?
Stuff is just very, it's
very physics oriented, right?
You know, if, if this temperature
goes up too high, then we know we need
to, you know, do these things right?
Denis: Mm-hmm.
Jim: A lot of companies don't
really have enough of that in place.
even know where to start layering in the
AI or machine learning stuff, because
they're like, they, they throw, they think
it's, it's the analogy of I have a hammer
and therefore everybody must have a nail.
Denis: Right.
Jim: And in some places,
like, no, I have a hex bolt.
No, I have a, a Phillips head screw.
No, I have, you know, an Allen bolt.
You can't just take a
hammer and pound those in.
Right.
And that's what I think
a lot of management does.
So our real concept is how can we
just take a broad view of all of
these things, you know, systems,
culture and people, data quality.
You know, what do you even
have in place today to figure
out where are your big holes?
Where are your big gaps?
And so that's kind of the thing
I think we're gonna build in the
coming year is how do we do a study
that's not super invasive, right?
But allows us to see these kind of things
and really go to leadership and tell them
the things that they don't wanna hear.
And probably what they're gonna do
is they're gonna crumple it up in a
paper and throw it over their shoulder
and go do what they want anyway.
But at least when that fails,
they're gonna have a plan, right?
I mean, they're gonna be
able to come back to it.
'cause somebody's gonna keep it and go,
you remember when we had these guys in,
I, I think they knew what they were doing.
Can we go talk to them
about maybe executing?
And realistically, some of
the stuff we're talking about
here, like it's not our space,
Denis: Yeah.
Jim: actually have a, have a customer
right now, or a potential customer
that we think we're gonna be able
to test this concept on, because.
It's a holding company and they've
got maybe a hundred manufacturing
plants in like 10 or 11 different
like divisions, so to speak, right?
Different.
of bucketed them and they're in all
different kind of places, right?
So how can we get them
somewhat level, right?
And so I think that'll be an,
an interesting case study.
You know, how can, and then how can
we take that other places, right?
Denis: Yeah, absolutely.
I mean, you mentioned management and.
I don't wanna bash on management in this
podcast, but I can't help but notice
the cognitive dissonance where you have
a higher up manager at the same time
complaining about not having the latest
number of this month for reporting,
but at the same time considering AI
to be the future of their business and
making monster investments into that.
Jim: Right.
Denis: So it seems we are
building on a crooked foundation.
Jim: Yeah.
that's, I.
I am just gonna say it.
I think we have a massive bubble with
AI because we have all this money.
I, I read something yesterday, um,
a guy named Ron Hetrick on LinkedIn.
He was saying something like,
71% of all venture capital
money is going to ai Right now.
Denis: Mm-hmm.
Jim: seen, know, I, I can't remember
whose study it, was it Yale,
Harvard, m, mit, I think it was MIT.
Denis: Yeah.
A big one.
Mm-hmm.
Jim: Yeah.
They said like 95% of all AI.
Um, projects are failing.
Like not, and we need to tweak it.
No, they're an abject miserable failure.
And yet you're hearing so many
companies laying people off
to go invest money in this.
You have all these data centers being
built, they can't even tie to the grid, so
they're having to bring their own power.
Right.
So it's, it's, it's a really
weird thing going on right now,
but no one or very few people.
Are actually able to tie value
to what they're doing with
artificial intelligence today.
Denis: Yeah, I'm yet to see anything.
Jim: now what I, what I will say is
I'm probably the only one on my staff
that doesn't use AI actively, where,
where my people use it, know, is for
instance, they're writing some code.
You know, like.
Justin was working on something
yesterday and we were pushing data.
We, he was working on flow, which is
something fairly new to us, and he
was pushing data out to a SQL backend
so he could do a Power BI report
on some downtime for this customer.
And what he then wanted to be able to do
was convert it to something else, like
maybe Postgres or something like that.
So he used AI to say,
okay, here's my SQL query.
How do I make it?
Into this other platform, whatever it was.
Does it work?
not sure yet,
Denis: Mm-hmm.
Jim: will save, you know, a ton
of time just by getting us close.
Right?
Denis: Yeah.
Jim: And of fact, Lisa from from
Group five sent me a blog other day.
She put one of my videos, my two
minute on Tuesday videos that I do
on LinkedIn into, I don't know if it
was chat, GPT or what, and said, Hey,
Denis: Mm-hmm.
Jim: based on this video.
Okay.
And she took a bunch of stuff
out of it, and then I read it.
She sent it to me and I read it, and I
could tell it was, you know, ai, because
I saw a bunch of M dashes in it, you know?
And I'm like,
Denis: That's a giveaway, isn't it?
Jim: yeah.
It's a dead giveaway.
All right.
And so I'm like, okay,
this, this isn't anywhere.
Well, it's, the concepts are good.
It's nowhere close to how I would say it,
Denis: Mm-hmm.
Jim: And so I, I rewrote it.
Realistically, it, it
took me a lot less time.
You know, and I probably
would've never written it, you
know, had she not done that.
Denis: Mm-hmm.
Jim: So it's not that it
doesn't have some news, right.
But think what people are trying
to do, they think it's gonna solve
problems that I'm not sure it's
capable of, and it's current form
Denis: I fully agree.
Jim: there.
You know, we got Altman and, and
Elon Musk saying, oh, we're gonna be,
you know, what is it they call it?
Um.
Sentient or whatever.
I mean, they, they're, they're saying,
oh yeah, we're gonna, it's, it is
gonna be able to do its own thing.
You know, like Elon Musk was like,
yeah, we we're up to about a 10% chance.
You know, that rock's
gonna be able to do that.
Sam Altman's like, oh,
we're less than a year away.
I heard that about Google a year ago.
If that's going on, like
it ain't gonna be public.
Like maybe our military and government
know how to use it, use it that
way, but we're not getting it.
And that's scary in and of itself.
But what Ron was pointing out though
was that despite all this in investment,
despite all these failures, no one
really has figured out how to build
new products, make new services, do
things that enhance people's lives
with this technology at some point.
These venture capitalists are gonna be
like, I'm not getting my money back.
That 'cause 'cause they're all,
they're all giving each other money.
Right?
It's kind of like, I don't know
if you ever watched The Three
Denis: Yeah, like a B scheme
or like a pyramid scheme.
Jim: it, it is, there was,
there was an old Three Stooges
get, you know, where like.
Curly owed mo, you know,
$20 or something like that.
And so like they end up passing
this thing around to each other.
It's like, well, here you
pay him and he, he pays you.
Denis: Mm-hmm.
Jim: like they do it.
And literally at the end of the
skit, they're all convinced they have
settled all their debts and all they
did was pass the same note around.
Right.
to me, that's what's going on here.
And so, you know, my concern.
You know is I'm, I'm saying Lynn, all
that happens, there's gonna be a kind of
a spirit of disillusionment out there.
What are you gonna do?
Because you tried to take the
Ozempic pill to lose the 30
pounds of weight that you want.
That didn't work, and it came with a lot
of side effects you didn't want, okay?
Now maybe you ought to go back
to what the personal trainer
told you two years ago, right?
That you need to get your
diet right and you need to.
You know, start working out, which
is exactly what the, the report and
the UDE score is gonna tell you.
And what I'm afraid of is in
today's environment, you know,
the executive's gonna ball it up
and throw it over their shoulder.
'cause they're like, that's hard.
That's gonna take time.
I don't have that.
I gotta do AI today because
Denis: Yeah, I mean, when it comes
to ai, I wouldn't say like I'm trying
to leverage it as much as I can,
and so far what I've been doing, for
example, it'll transcribe this podcast.
It'll maybe correct the audio, it'll
correct some of my emails and the way
I see it, it essentially allows me
to spend more time in my genius zone.
Jim: Mm-hmm.
Denis: say a New York lawyer
can spend more time billing
clients for difficult work.
In no way, shape or form do I see AI
doing the stuff that we are talking about.
For example, assessing ude.
So I would pay $100 per month for ai.
I would not pay $200, and
apparently the cost per customer
at this time at open is $200.
The method doesn't check out.
There's some benefit, but nowhere
near to warrant the number of
VC money it's raising right now.
Jim: Yeah.
And I, and, and I think there's, I
think what he was saying too is money
being poured into this, it's 17 times
What went into the.com
bubble, IE around 2000, 2001,
and four times what we saw in
2008 with that real estate market
and the, the subprime mortgages.
Right.
Denis: Right.
Jim: lots,
Denis: Mm-hmm.
Jim: So it's like an even bigger bubble.
Right.
And how long does it last before?
You know, people realize maybe this
isn't what it's cracked up to be.
And I, I talked to a customer, uh, maybe
a couple weeks ago and I, I can't, I can't
talk about what industry or who it is
or anything like that, but they actually
were doing some really cool stuff with ai.
They were empowering their people with it.
Right.
They're, they were allowing, 'cause
like maybe they have a central
development team that's building a
lot of their major solutions, right.
Well the, the issue, and this is kind
of where we fit in a lot of times,
helping people, helping customers
out is, you know, maybe the central
team's taking care of the big
hitters, but there's, you know, maybe
they're hitting the home runs, right?
If we're going staying in sports
analogies, man, there's a lot of
singles and doubles out there, right?
They can really score a lot of
runs and those are kind of being
ignored 'cause you're trying to
go for the big home runs, right?
With the central group.
Denis: Mm-hmm.
Jim: And he was telling me, he said,
yeah, basically we have some internal
stuff we've built allow some of
our users like build some of these
solutions that we can't touch as a
central team 'cause it's not worth it,
but that are worth something, right?
It solves problems for them and
a few people around them that's
actually worth money to the company.
And he said that's kind of
how we're using it versus some
whizzbang overarching thing.
We're just trying to figure out how to
make people more productive and, and more,
you know, help them do their job better.
And I'm like, that's probably the
right approach and that's not anyone
else's doing or very many are doing.
I thought it was a very prudent.
But I thought the concept in the
Denis: Okay.
Jim: of that was, was probably right, so
Denis: makes sense.
But let's return to before 2022
when AI was still machine learning.
We still had computer vision and
so on before it all became LLMs
Jim: Right.
Denis: to me, feels like we were
trying, we were working on that.
We had some progress, but the
progress was slow due to well bad
data and basically not being ready.
Now, instead of fixing that foundation,
we just now piled up LMS and
completely forgot about the problem,
and now want to do an even bigger
challenge with the same problems.
So if we now return to, well,
what the current problems are,
we try to measure with attitude.
Let's say we are, you have someone
listening here who owns a plant
with, thinks want to leverage
AI or VR or blockchain or even
classical machine learning,
Jim: Mm-hmm.
Denis: is unsure of whether
they can support it or not.
What would be a good place to start?
What systems would you
have this person examine?
Is it something more cultural?
Technological?
Jim: So the way we've machine
learning, you know, even in a.
A plant or a place that doesn't have
a lot of the basics in place, per se,
Denis: Mm-hmm.
Jim: people to at least try it, right?
It's not, it's not gonna
be the be all, end all.
What it will do is it'll show you
gaps, and it also allows us to build
out some of that infrastructure.
Like we have a chemical
manufacturer, the one that I was
doing the training curriculum for.
We've done a bunch of machine
learning work for them with Simco.
Denis: Mm-hmm.
Jim: As a matter of fact, the first
machine learning project I ever was a
part of, and this was in 2018 or 19,
it was fall of 18, is when it was,
was an absolute miserable failure.
It was six weeks of just,
this is never gonna work.
It was the wrong tool, wrong
algorithm, wrong company.
Who didn't understand the
problem unfortunately,
what actually ended up happening, we
built, we built a first principles set
of analytics to put on a dashboard in
front of their operators where they
can make a more informed decision, and
it has saved them millions of dollars.
It has literally
Denis: Mm-hmm.
Jim: successful from a dollar
perspective project we've ever done.
Was it wrong that we
tried machine learning?
Absolutely not.
learned a lot.
Right.
It was a failure and we've
done a lot of others.
We, we do a lot of stuff with Simco.
They have a lot of batch processes
and a lot of times like, Hey, we
were running really good two or
three years ago and now we're not.
Or, Hey, we have this quality problem
and is the one quality metric.
We've never really figured
out why it happens.
Generally, we kind of know if this
quality metric goes bad, here are
the three or four things we look at.
one we don't understand, and so.
What was, in some cases we ended up
having to build some infrastructure
to get the data lined up where
we could do that analysis.
And what I tell people when we start
doing machine learning projects is
sometimes we get no problem found
Denis: Hmm.
Jim: and, and I'll never forget the
one that really, really taught me this.
Was, this was for, um, it was
for the old Verso paper plant
up in Anders Co in Maine.
Before it, they blew up their
digester so I can talk about 'em.
Um, we had just learned sim ICA and
come back and I was like, man, I really
wanna like, try to go solve a problem.
And so I messaged the operations
manager up there and I said, Joe,
I said, we just learned this new
tool and I want some practice.
And you know, we're not
super busy at the moment.
Do you have a problem?
We could go try this out on, if
we find something, we'll work
out some financial compensation.
If we don't find anything,
you owe me nothing.
I just, I just want the practice.
Right?
Or want us to get the practice.
He's like, great.
He goes, we have this apparent density
problem and I can't remember what a
parent density is on the paper machine.
There's it, it's a quality metric.
Denis: Mm-hmm.
Jim: And he said, yeah, we had this
step ch you know, we had this change
happen right around this particular
date, and you could do a q some chart
and see like, yep, that was the bottom
Denis: Something happens.
Jim: going up ever since.
Right.
And you're like, okay, yeah,
this is definitely a problem.
So I use Simco and, and it's kind of,
that's kind of a classical case, right?
Is you, you get data for
like that point onward.
Versus that point rear word and
say what's mathematically different
between these two data sets?
Right?
You gotta line everything
up the right way.
And I told him is I don't see
anything on the paper machine itself.
You haven't changed anything,
you know, reasonable.
'cause I was lining up
grade run versus grade run,
Denis: Right.
Jim: here's, here's this grade that you
ran and the apparent density was good.
that same grade run or set
of grade runs with it bad.
they're reasonably close.
There's nothing like, there's nothing
that shifted at the same time.
So then we started going to the pulp
mill and I told them from, from the
jump, I said, look, if I find something
and you're like, I can't control that,
or that is, that's correlative, but
it's not, CAU couldn't be causal,
Denis: Right.
Jim: it down the model.
Right?
Like you, you're allowed to call bs.
Like, I'm not gonna be that guy
that's shoving stuff down your throat.
Like a lot of data scientists who don't.
the process or don't try
to understand the process.
Right.
You gotta help me with that part, right?
And, and you're not gonna insult me.
So, you know, kind of, you know, we,
we went through it, went back to the
pulp mill, looked in the pulp mill.
Did anything shift over that time?
No.
So we didn't find anything, but then
I started asking him, well, what about
things that you're not collecting?
So for instance, you have a big pile
of wood chips in a paper mill, right?
Okay.
The stuff you're cooking in your
digester, are they from the top of
the pile or the bottom of the pile?
It's been rained on and sitting at
the bottom of the pile for 30 days.
They'll cook different.
Right.
I'm not a paper maker,
Denis: Mm-hmm.
Jim: I know they'll cook different.
Right.
Yeah, they definitely cook different.
Well, do you know which
one you were doing went No.
What about like on the paper machine?
Do you know if you've changed, like
felt supplier or wire supplier or?
Chemistry or anything like that.
Didn't have data on all this stuff
or couldn't get to it easily.
IE Why we go to ERP and MES
and other systems, right?
Because hey, maybe they clo they
changed their clothing supplier or
felt supplier on the paper machine
right before this QSO thing happened.
We couldn't get to that data.
Easy to start correlating it.
Denis: Mm-hmm.
Jim: we got to the end of that.
And we literally had no problem found.
There was nothing in the data that
they were collecting that told
us here's why the shift happened.
What I was convinced is it was
data that they weren't collecting
or that we couldn't get to.
That was really the problem.
Denis: Yeah.
Jim: now all of a sudden you
start layering AI on this, right?
And you kind of have the same problem.
So what we try to encourage customers to
do that's not a worth worthless endeavor.
It teaches them here, here's the kind
of stuff we're gonna have to have in
play if we want to do this quickly,
they, we show them all the groundwork
we're having to build to get to that,
right, to even be able to do this.
Here's all the stuff I had to
do with your data to make it
ready for something like this.
So if you want to be able
to scale something out.
gotta do this in spades, right?
So I always encourage people, let,
let's go give it a try, right?
Let's, let's go what we can do.
You know, with, with, uh, the LLMs,
I, I'm a big fan of, why don't we do
small, small language learning models?
don't we take all your documentation, all
your, you know, trouble cause correction,
documents all of your operator documents,
you know, how you run your facility.
We could easily build an LLM or a
s uh, SLM, small language model,
know, and make a, a teams chatbot.
One, a couple of my guys did
that for some PI stuff, right?
How to upgrade a PI system.
They took seven or eight, you know, um,
PI system documents, did, did training
on it, built a chat bot in teams, and
now I can ask that, Hey, I'm going
from this version to that version.
What do I need to do?
And it'll give me the list of instructions
and cite where in each manual it found it.
Denis: I mean, that's great, right?
And even regardless of outcome, in
this case it worked, but even if
it doesn't work, that's one thing
I appreciate about this AI wave is
that it really forces companies to.
Do more with their data to collect data.
'cause I dunno how it was in the
US but I'm mostly based in Europe
working for German companies.
And up to not so long ago, there
was really an attitude like, yeah,
but why do we need this data for?
And at least now they fool themselves
into, well we need data for ai.
Whether AI will work out
or not remains to be seen.
But hopefully after a couple of
years at least they would have
a proper data infrastructure.
Jim: Yeah, I, I think AI is
gonna play a part, right?
It's not gonna be as big a part as
what a lot of people think, right?
You just a tool in the tool belt.
It is not a be all end all.
I mean, just five years ago everybody
was saying machine learning was gonna
do all this and it didn't either.
Right.
It's a
Denis: And before that was big data.
Jim: Yeah.
And big data.
Yeah.
We, you know, digital twin, like, you
know, throw all these, these terms out.
Right.
They just keep, you know,
kind of moving the goalposts.
Right.
But the reality is, it's, it's just like
when we were talking off camera, right?
You can't go.
run a marathon and you've been
sitting on the couch eating
chips for the last year, right?
You can't just get off the couch
and go run 26 miles tomorrow.
Right?
There's a process there.
There's base level things you
have to build to do that, right?
I've been on a fitness journey for
the last four year, five years.
Right.
I'm about to turn 55.
I hurt myself playing golf of all things.
I I got a bulging disc in my neck
from hitting so many golf balls.
'cause I was trying to break
70 by the time I was 50.
Denis: Mm-hmm.
Jim: And what happened is I had a weak
left shoulder from an old injury, an
old, um, torn rotator cuff injury that
finally reflected back into my neck.
And I'm like, I hurt myself
playing golf of all things.
And so I was like, well, that's
never gonna happen again.
And so I started, you know, working
out, you know, and consistently.
And the thing is though, like some
of the stuff I do now, there's no way
I could have done now or could have
done then what I can do now, right?
Because I have four years or
five years worth of base, you
know, to work from, right?
So I can do more challenging things.
And I think a lot of companies,
they just, they still because of the
culture that we all live in, right?
The investment culture, you know,
here in the states, the Wall Street
culture, it's, you know, what
have you done for me late lately?
What
Denis: Quarterly profits.
Jim: Yeah, it's, it's,
it's all quarterly, right?
So it's like, you only made 1.1
bajillion dollars.
Last or last quarter you
were supposed to make 1.2.
So therefore we're selling, right?
And then all of a sudden their stock
plummets you're like, well, didn't
they still make a bunch of money?
Aren't they going the right direction?
Yeah, but they don't know their business.
They only made 1.1
bajillion.
They should have made 1.2,
like they said.
And what it, what ends up happening is
we get all this short term thinking.
We get a lot of fear of missing
out stuff going on, and we
get the, the situation where.
People want AI now.
Yeah.
Denis: All right, so there was
a small hiccup in my connection.
Apologies for that.
Jim was just explaining about the
short term vision of the industry.
Jim: Yeah.
And so what I was, what I was saying
when Dennis like got jettison, there
was, um, you know, these companies want
AI now, Because they have this very
short term quarterly kind of a focus.
And, you know, I was, I was saying about
this during the break, Simon Sinek.
You know, he wrote a, I think it was
a book he wrote called The Infinite
Game, and the concept is, is with
companies, what you're really trying
to do is just survive long term, right?
There is no winning, there is no losing,
there is no like scoreboard, right?
Like baseball, you play nine
innings, there's a scoreboard.
Football, you play 15 minute quarters.
There's a
Denis: Mm-hmm.
Jim: There's a way to keep score.
There's a defined end to the game.
There's a defined rules
to the game, right?
With companies, it's really not.
Right.
It's an infinite thing.
You're trying to, you know, stay
in business, survive and, and
grow and thrive and all that.
Right?
And I think so many companies are driven
by, they think it's a finite game, right?
And realistically, the things we're
looking at is very much why, it's
why I use the analogy all the time.
It's the fitness analogy, right?
It's not like.
Okay, I can bench press 500
pounds or I can run a marathon
in under two and a half hours.
I won.
You're doing it more for longevity
and quality of life, It's not a game
that you win, like, oh, I mean, yes.
You set goals along the way, for sure.
Right.
I mean, like first half marathon I ran,
I wanted to run in under two hours.
I did.
Right before I strained a calf, I felt
like I was gonna run a half marathon.
I was gonna run the St.
Jude in Memphis.
I thought I was targeting 1 45.
Now, did that mean I really won anything?
No.
I was very slow compared to the
fastest runners in the race, but
it would've been a win for me.
Right.
And why was I even doing all that
was to try to stay in shape because I
Denis: Mm-hmm.
Jim: Right?
That's really, you know, and so I
was able to like put a goal with
something really I'm trying to
do for my long-term longevity.
And I think a lot of the stuff we're
talking about today needs to be
looked at, you know, differently.
This is a long-term investment you're
trying to make, and too many companies
are like, well, I wanna do this now.
I want to do it yesterday.
I want to, you know.
You know, and here, here's the big issue
I see is it feels like what's going
on is a lot of companies are trying to
replace people with these kind of systems.
And maybe some of it is we have
boomers retiring and we don't have
as big of a population coming in.
Right?
That may be part of it, but it feels like
they're even trying to take the place of
the ones that are coming in to replace.
So, okay.
Every company lays off
25% of its workforce.
what?
buy your products?
'cause we're gonna have a crashed economy.
Denis: Mm-hmm.
Jim: So one of the reasons why I just try
to focus on a lot of the basics is because
even if you know the technologies change
in advance, you still have to have this
foundational layer to be able to apply
them and use them no matter what it is.
Whether it's an LLM, machine learning or
something, quantum computing, whatever
goes with that, you're still gonna have
to have a solid foundation and a solid
base of quality data, accessible to you
in a way that people can do something with
it or systems can do something with it.
And almost no company
Denis: Yeah,
I mean, to be fair, the definition
of having a, like what is a
good digital infrastructure?
It seems that after decades we are.
Fully converged on one idea, yet.
I saw a lot about the
Jim: No.
Denis: but Ifda shortcomings, right?
Especially in the process industry.
It's mostly time series.
I kind of get
Jim: Yeah.
Denis: say, leaders who are
confused, like, okay, fair enough.
But then what should we build?
What can you tell about that?
Jim: Yeah.
And.
Yeah.
And, and that's one of the things,
there's, um, I wanna say it's Ryan Hill.
He had written this book about,
you know, how to leverage AI in
a modern manufacturing space.
And, you know, he was talking about, you
know, you have to have, it's not really
unified name space, but common data
Denis: Mm-hmm.
Jim: right?
And so I sent him a message.
I'm like, okay, where can I go
build this common data model today?
Like, what product can I go
buy that allows me to do that?
And he kind of admitted there
really isn't one they've, he's built
components of it with some of the
tools he's used over the years.
I could say the same thing, right?
With some of the PI asset framework stuff
we've done, but it's only a part of it.
So there's nobody really today that I
see that allows you, and I think there's
some people working on it for sure.
I mean that, that to me is where.
You know, coming in like flow,
coming in, like high Byte or litmus
automation, they're trying to you
to build these type of data models
and let the data stay where it's at.
Denis: Right.
Jim: Right.
Because one of the shortcomings with,
with the PI system, you know, we were
kind of talking about this yesterday
when Justin was showing us everything
he had done with Flow for this customer.
I was kind of asking him and Ben like,
what do you guys really think of this?
Because here's how I think of it.
And I'm like, you know, we
go in the right direction.
And, and they're like, yeah, I mean,
'cause it's like really difficult to
do some of this stuff we're doing here.
Like, yeah, could it be better?
Like, yeah, but they, they recognize
Denis: Hmm.
Jim: re need to rewrite this,
know, it needs to be easier.
But they're like, we can't do some of
this stuff with any other technology
because one of the problems we see, like.
With, 'cause what, what they, what Justin
was able to do was, wanna pull data.
They had an ignition historian and they
had, um, an MES system that was homegrown,
I think it was homegrown, it was SQL
Denis: Right.
Jim: And so he had to grab data from
both of those, put them in a hierarchical
structure, do some calculations, and push
the results of those out to power bi,
like looking at downtime for instance.
'cause they couldn't
have one way to do that.
So.
You know the thing is that like, could
we have done that in say, a pie system?
Well, the problem with pie is
you have to bring everything into
pie to use those kind of tools.
It can't bring data from multiple sources.
Right?
And so I think that's where why,
what, you know, flow, litmus, high
byte, and some others are trying to
build is actually pretty important.
'cause you can leave the data where
it is, but then build this model.
You know, to kind of organize the
data, just Plantwide, but potentially
companywide enterprise wide, right?
Because I don't think there's
really a product to do it today.
Denis: No, there's some things
moving, but for example, edge Lake
is an open source data lake that
keeps everything located where it is.
Um, but even that, it's also new.
I don't, I don't know any
finished solution either.
Even Databricks, which has been growing
tremendously, still needs all the data in
Jim: Yep.
Denis: which is a huge
journey to complete.
Jim: It's a, it's a big journey to
complete and it's gotta land in context
or you've gotta be able to contextualize
Denis: Yep,
Jim: right?
Because, 'cause the thing is, you
know, if you go look at, say, uh, you
know, in data historian really of any
kind, more times not, they've named
the tags in such a way that only the
process control people understand it.
Denis: you're right.
Jim: Somebody counting
isn't gonna understand it.
Like, yeah, that's the totalized
flow and tells you what you
produced out of this particular unit
Denis: Mm-hmm.
Jim: but it may or may not be named such,
you know, it's FQI 1 23 point pv, and
it's like, okay, the process engineer's
like, okay, FQI means it's a flow and
it's a totalized indicator, right?
But then they have to know,
well, what does 1, 2, 3 mean?
Right?
Oh, well, let me get the p
and ID drawings out if I don't
know that area of the facility.
And, and what's going on is get
higher up the food chain, know, in,
especially if the people in those
seats didn't come up through the
ranks, they have no idea that that's
really how things work under the hood.
Denis: Yeah.
Jim: They have no idea.
Right.
You get somebody who's, you know, CEO
level and they started out in one of the
facilities and worked their way up, got a
Denis: Mm-hmm.
Jim: But a lot of those folks, aren't
they, you know, a lot of people in the CO
roles didn't necessarily come from that
Denis: No, not anymore.
Jim: Don't necessarily understand
the industry, especially not.
At a deep level, right?
They're just business people.
They have MBAs and know
how to move things around.
Closed plants, you know,
consolidate things and not lay
Denis: Mm-hmm.
Jim: being very here, but know,
I'm not trying to beat them up.
But they don't have that deep level
of knowledge of how does, how do all
of their manufacturing facilities
even work and what's really important.
And yet they're making
these big decisions on what.
We need to do without something
like, can we even do it in place?
And depending on culture, you may
have a situation where you have a
bunch of yes men telling us, yes,
we can go do that, no problem.
And then they go yell at people
down the chain like, why isn't this
Denis: Yeah.
Why isn't it Mm-hmm.
Jim: Yeah.
It's like you guys have no idea.
Right?
And that's, I think, you know, to
kind of tie it back together, I think
that's why there's the idea of this
UDE score is, is important, or UDE
score is important because we're trying
to give these companies a realistic
assessment of where they're at in their
journey, where are their shortcomings,
where are their holes, you know?
And realistically, we're not gonna be
the company that fulfills many of these
Denis: Hmm.
Jim: They may have massive holes in aqua.
Matter of fact, we, we were helping,
um, a company, fairly small facility,
part of a larger organization.
don't have a limb system.
Denis: Right.
Jim: We built kind of a poor, poor
person's limb system in their PI
Denis: Mm-hmm.
Jim: and it took way longer
because PI's not a limb
Denis: Mm-hmm.
Jim: you know?
And is it ideal?
No.
But we're trying to help them
justify, Hey, we need to be
able to have one of these.
They were keeping it in a spreadsheet
and then they couldn't equate things
and weren't able to catch problems
before they got out the door.
Right.
Well, now they're gonna be able
to see things a lot more clearly.
You know, the reality is, you
know, sometimes companies don't
even realize, like that's the very
thing that is holding them back.
And so like if, if they come to us and
say, Hey, we need to put in a limb system.
Okay, we'll go talk to some of our
customers on what, what they've
done, but we're not the ones to
put that in and manage it for you.
That's not our area of expertise.
But what we need to be able to do is
help them identify like, Hey, this is a
really big gap compared to other companies
like yours, this is how they use it,
and this is how they use it to drive
better quality products out the door.
First time.
So it might, might not even be
things we even touch, that's okay.
That's okay.
With us, it's, it's really about
how can we show people here are the
gaps you have in the things that
you really want to get done and
maybe reduce some of
the fear of missing out.
You know, I, I would love for, some
customers are doing this, I would love
for some of them to say, you know what?
Yeah, I know AI is all the rage, but
we're not even close to ready for that.
Let's get some of this,
this fundamental work
Denis: Mm-hmm.
Jim: We're gonna be paying
ourselves, you know, along the way.
Yeah.
We'll experiment with some AI things like
maybe we'll take, you know, for instance,
I think one of the easiest barrier
entries is take their documentation
and put it in a small language learning
model and then they can claim, Hey
look, we're, see we're using ai.
Look, it's helping our
Denis: Yeah, that's a hanging fruit.
Jim: you know?
Yeah, it's low hanging, right?
It's, it's not some whizzbang thing that's
gonna, you know, cost you a ton of money.
Like we, we literally have a food
customer with, with us right now
who did a big project with one of
the, and they have AI in the name of
Denis: Mm-hmm.
Jim: and it was a miserable failure.
And they're like, you know what?
We need to go back to basics.
We need to have all these dashboards.
We need to understand and
be able to educate people.
how to use data the right way first,
and then maybe we can pursue that.
And we kept trying to
tell 'em the whole time,
Denis: Yeah.
Jim: wouldn't listen.
Denis: But companies have to
play the game as well, right?
For example, at my current
client, budget is being cut for
maintenance of an existing system.
But then again, there is more budget
available for a new big AI project.
So now some people are staffed under that.
So it's basically, you put it from
one pocket to the other pocket.
But unfortunately, managers also have to
play understand that AI is the big storm.
Now, that's the way to get
your, um, project funded.
So in some sense, they have to
banner to the top, so to speak.
Jim: Well, and, and, and the thing
is, is if you don't have an AI
strategy, if you're a publicly
Denis: Mm.
Yeah.
Jim: right, know the, the stock markets
Denis: It's gonna punish you.
Yeah.
Jim: right?
They're gonna punish you.
They're not gonna invest.
Oh, you're not investing, you
know, 14 bajillion dollars in ai.
Well.
You know, maybe we should sell our
investment in you 'cause you're not gonna
be here when it may be the most prudent
decision, you know, they've ever made.
But do you think a CEO's
got the courage to do that?
No.
'cause the board of directors will
Denis: Yeah,
Jim: him.
Denis: He's not omni potent.
Jim: And, we're, we're in
that, I mean, our culture so
much, this fear of missing out.
Right.
You know, we, we.
Everybody wants to keep
up with everybody else.
And you know, like my wife and
I were talking over dinner last
night and she's like, I just don't
understand why we're we're doing this.
And I said, well, here's the reality,
know, is if one country's not pursuing
it, like let's say the United States
isn't pursuing this, these kind
of technologies, well we know our
enemies China and run Russia are
Fear missing out.
And I said it's the
same for every company.
You know, it's like if I'm x, y, Z
paper company and I'm not pursuing
this, then you know, A, B, C and DEF
paper companies are doing it, and
if they do it, they may pass me and
leave me in the dust, so I've gotta go
Denis: Mm-hmm.
Jim: It's, it's definitely this
culture of fear of missing out and I'm,
I don't, I don't know that it ends
Denis: Yeah, I mean.
Y.
Jim: I really don't.
Denis: fair.
I mean, I'm not against trying to invest
in AI and getting it to work because there
is a small chance it may pay off big.
I think what you and I are against
is like doing it too quickly and
not doing the foundational work.
Skipping all the, getting
healthy, cleaning up your
diet, exercising regularly.
It's, as you say, getting off the
couch from eating your bag of chips to
running a marathon and then spraying
an ankle and make an expensive mistake.
Jim: Right.
Or worse.
Denis: Or even worse,
Jim: Or worse.
Denis: the summer Thought
Jim: Yeah.
Denis: just to finish it about, um,
you mentioned people and culture
and you mentioned indeed that lot of
companies try to leverage or make AI the
thing that is gonna reduce headcount.
Not only senior people, but also
unfortunately people coming in and
especially Gen Z, really feels this.
And I think it's so sad because a
lot of the power of analytics is
when you have experienced people who
are passionate about the data, have
business knowledge, have experienced
use it and build something with it.
And it feels that we
Jim: Yes.
Denis: that very thing as a
lot of value to this industry.
that also how you see it?
Jim: Absolutely.
I see it that way.
And, that's where, I mean, I, I feel like
we kind of fill a little bit of a gap,
you know, there, because what I'm seeing
with so many companies is they've lost so
much experience over the last Oh, really?
Since COVID so many,
you know, boomers have.
Either through layoffs or just done.
You know, they see the writing on the
wall and they're like, you know what?
I'm taking my toys and
Denis: Mm-hmm.
Mm-hmm.
Jim: And the, the problem we're
seeing is what's replacing them.
This is the typical scenario, right?
So let's say senior process
engineer at at plant A says that,
Hey, I'm hanging up my cleats in a
Denis: Mm-hmm.
Jim: Right.
Oh, we better extract all the
knowledge we can out of them.
They hire some young gun
engineer, young gun engineer works
under, this, this person, okay.
That goes well.
They're not gonna pour, I mean,
it's like trying to pour a
five gallon bucket into a thi.
Doesn't matter how fast you pour,
only so much is gonna get in in a
Denis: Yeah.
Jim: You can't take 25, 40 years of
knowledge, especially if you're not
trying to build digital tools that
are really capturing this knowledge.
Right.
Nobody really thinks of that.
And what, then ends up happening is
now this, you know, the engineer, the
old experienced engineer goes away,
no interest in back in or, or is like,
yeah, I'll work 10 hours a week or
Denis: Mm-hmm.
Jim: You know, at twice, right.
They were making, or three times
now, this young engineer.
All of a sudden has, has
done some really cool things.
Well, they either get promoted
within their own company, one or two
worse, they get recruited outside
Denis: Mm-hmm.
Jim: and they leave.
Now, not only have you lost the
experienced person, but your young
person who learned from them is
Denis: Yeah.
Jim: Now you hire the next
person in who has no experience.
And, and no download and you haven't
billed systems that help them
troubleshoot what the other two knew.
That's where we're at.
And, and it's, it's, um, was a guy
probably seven or eight years ago was,
was talking to me about this and he said,
Jim, he goes, we have to do all of this.
We have 'cause because he was
seeing the writing on the wall.
We have all these people
retiring, he said.
We have to build these kind of systems
before the emperor has no clothes,
and he is about half naked now.
And I made a statement
at Aviva World last year.
I said, well, if, if he was half
naked seven or eight years ago, he is
down to his underwear and skis now.
You know, and skis, this
is a, a funny name for
Denis: Right.
Jim: right?
Funny old slang, word for underwear.
And so I'm like, we're almost at a
point where you can't get to that.
And we're seeing it one
at one of our facilities.
We're we're doing some.
Multi-variate work, trying to figure
out what's going on in a batch process.
Who's driving it?
engineer who's getting paid
probably twice what, or three
times what he used to get paid.
But he's only interested
in working so much.
Denis: Yeah, of course.
Jim: he's fishing.
like, I'm, I'm done with this.
You know, I'm, no, I'm not
traveling to the plant.
I've moved up here.
I'm on the, this river and I'm
going trout fishing regularly.
Right.
I'm
Denis: Yeah, which is fair enough, right?
Jim: And Yeah, I, I mean, he's earned it.
And the thing is, it's, it's
interesting is we cannot get the
young engineers to dig into it.
They're like, ah, I don't need that.
You know, I don't have time for that.
I'm too busy fighting fires.
And we're like, dude, if you don't
have this, like you have no chance.
like, we cannot get them
interested in looking at it.
And what's, what's interesting is.
When this fellow was at the facility,
the first time I ever met him, I
was showing him PI event frames, and
I'd thrown some stuff into Power BI
to allow them to compare batches.
And he's like, well, why do I need that?
Because I have PI Batch view.
See, I have all the views.
And I said, let's see.
I'll call him Charles.
His name's not Charles.
Call him Charles.
Charles I, I'm not building
these solutions for you.
I'm building it for who comes behind you.
Denis: Yeah.
Jim: I gotta give them a chance.
I gotta, I gotta get all your knowledge
out and out of that into something
like this where they have a chance.
And what was interesting we did
some, some Simco work for him about
three or four years ago where I
think I, yeah, I was doing that work.
And I found the same stuff he found,
I had a, I had a decent
amount of the foundation.
I had to build some other analytics.
For some of the things he was,
he was thinking about that
he, he was seeing going on.
I built those, put them in
the event frames, and then
used that in my sim analysis.
Took me about three days and
I'm like, here's what I see.
You know, I see these things are the
reason why your batches are running slower
at times than you know, than you're not.
And he said.
Well, you just figured out in
three days what I've been staring
at for three weeks and I've got
30 years experience and you don't.
Denis: Hmm.
Jim: So it was nice that it validated
what, what each other was seeing.
Right?
He was glad to see like, hey, I
used a modern tool to validate his
thought process and I was glad that
I was able to take his 30 years of
knowledge and his three, three weeks
of looking that validated what I found.
And as I told you, I, we also find
a lot of no problem founds, right?
So we don't always find something.
I would say about half the time
we're reasonably successful.
Maybe 10 to 20% of the time
we'd find something like
this where it's a slam dunk.
But it was really nice, you know,
that we were able to find that.
And he was literally a
year from retirement.
So he's a big fan of the stuff.
was the one that called us about
this particular process as a retiree.
Saying, they're asking me to look at this.
I know this tool you have works.
Help me figure this out faster.
Denis: Yeah.
Jim: Right?
And so it was nice to kind of see that,
that, Hey, here's this, this guy that
typically is the, well, I've always
done it this way, I don't need that.
Is now kind of like, Hey, I
actually see the value in this.
We can't get the young
person to look at it.
How weird is that?
Right?
But, but I think the, the thing
we see pretty consistently is that
a lot of retirees are, are, are
just, they're, they're gone.
And the, the people that have replaced
them don't have nearly the experience.
So A actually, it makes it even harder.
You don't have the foundation,
you don't have the institutional
Denis: Yeah.
Jim: were talking about.
To where we can use that institutional
knowledge with these great tools to
solve a problem better, faster, and
put it to bed completely forever.
You know?
And now you have very inexperienced
people who are like, well, I, I
don't even know good even looks like.
Denis: It is a big problem.
No, I fully agree.
And that, in that sense, I see AI
as a risk, but if you would take,
let's say, the key points of this
conversation and give advice to,
let's say, a plan manager listening.
I think from our discussion
we can definitely say that.
Fine, if you want to risk.
Experiment with ai, go ahead and
do it, especially if, uh, gives you
more budget for your important work.
But in doing that,
Jim: Right.
Denis: forget to build your
solid foundations that we
discussed under your attitude.
Namely network systems,
data and also culture.
Jim: Culture.
Denis: in terms of
Jim: Yeah.
Denis: makes sense that you don't
replace the young people because
you're gonna need them anyway.
You won't have your boomers
Jim: Yes.
Denis: people forever.
Is anything you would
add to this list, Jim?
Jim: No, and I, I tell 'em just to
take a balanced approach, right?
So yeah, we can do this AI thing, but
just realize that 70% of that effort or
Denis: Mm-hmm.
Jim: your foundation, right?
And let's be very smart at what we target.
Right if, and be very
specific at what we target.
You know, when we do some of the
machine learning stuff, like we've, I,
I, we've counseled people for years.
Now let's be very specific.
Let's not take thousands of
variables and a very broad problem.
No, let's get very specific,
Denis: hanging
Jim: you know, be
Denis: simple.
Jim: because, because it's gonna
get really muddy that, I mean,
that's what happens when you
throw a lot of data at something.
It just gets muddy.
Denis: Yeah.
Jim: You know, and all of a
sudden these correlations, matter
of fact, I have a funny slide.
You know, where it, it'll be something
like the amount of milk consumed in
North Carolina versus the number of
lawyers in Australia, and they're
like a one-to-one correlation.
They have zero to do with, right?
That's the kind of stuff that
shows up when you, when you
build out massive things, right?
So I just tell people to be
very targeted and focused.
I'm still of the opinion, like there's
two ways you can do these, these
projects, you identify a very common
problem across all your facilities.
Like if you're at the corporate level,
and I I, I learned this from Jonathan
Alexander at, at Albemarle, right?
He's got a small team.
He had to go after a very similar
problem across every facility,
Denis: Yeah.
Jim: right?
Like I have to kind of solve
the same problem 'cause I have a
Denis: Mm-hmm.
Jim: If I want to get budget
and get attention and get what I
need done, I have to go after a
problem that everybody kind of has.
Like I have to take a little broader
brush and know, go after something.
Right?
Whereas we are much more
specific and narrow and so like
you can do it one of two ways.
Like if it's a plant manager,
I'm like, be specific and narrow.
Let's go very deep with
a technology stack.
Right?
Let's.
Let's build all the foundations.
Let's, you know, build first principles,
dashboards, let's build, you know,
uh, business intelligence tools.
Let's use some of the advanced
analytics tools, machine learning,
ai, whatever we want to choose.
Let's use it all on that problem.
And then you'll kind of see like what.
What needs to be applied where like,
oh, here we have this problem over here.
We can just do this base level
stuff and it'll solve that problem.
But hey, here's this problem that's
been plaguing us for 20 years and we,
we never have figured it out despite
having good first principle stuff.
Maybe we need to throw these
Denis: Mm-hmm.
Jim: that.
Yes.
You know, where they kind of
learn like, where can I apply
the right tool at the right time?
So it kind of depends on like
the lens plant manager narrow.
Go deep, right?
Corporate level, know, you
probably have to be a little bit
more broad with your thinking.
Like what is a common problem that this
kind of technology could help us with?
To me, like the whole documentation
and training, you know,
operators to do things better.
Like, to me, that's a perfect AI
use case because language models do
Denis: All right.
Jim: They can scan documentation
in the written word and summarize
it for us and give us curated
content back when we prompt it the
Denis: Mm-hmm.
Jim: That's proven right
you give it crappy data.
I mean, we, I, I actually saw
this and, and here, here's where
Denis: Mm-hmm.
Jim: a potential problem could be, is.
These, these models are, are very
confident, even if they give you the wrong
Denis: Yeah.
Jim: And I'll give you an example.
I shot a video about this, it was
a two minute on Tuesday video, and
I have the video evidence that this
happened because I didn't believe it.
Somebody posted on LinkedIn and I didn't
believe it, so I went out and did it.
So I, and I shot a video on it.
But the simple question was, is does
water freeze at 26 degrees Fahrenheit?
And Google's AI said, no, it does not.
It freezes at 32, therefore
at 26 degrees it's liquid.
And it doubled down on that answer
Denis: Yeah.
Jim: It said it twice.
It was a very, very
confident, wrong answer.
So if you give an operator a very
confident, wrong answer, it can lead
to a very serious industrial accident.
And I think that may have ha.
My prediction is that'll happen and that
will put the brakes on a lot of this.
Because if you give somebody a very
confident, wrong answer and then
we have a really bad industrial
accident over it, like then people
are finally gonna start looking at
Denis: Yeah, but it'll be too by then.
Mm-hmm.
Jim: right.
But if you have good documentation,
right, this could be actually be super
powerful where the operator's like, Hey,
this is what's going on in the process.
They ask a chat bot, what do I do?
Like I see this pot
potential thing happening.
Or even better yet.
You know, the process is starting
to recognize, hey, the, we've
seen this pattern before.
Here's what's likely to happen and here
are the knobs you need to twist, and
here's how hard you need to twist them.
Right?
Where, where it's a little bit easier
for them and, and maybe it's, maybe it
doesn't do it for you because, in some
cases you can, but you know, 'cause
I've been doing that with neural nets
for a long time, especially like in
the oil and gas space, we see a lot
of advanced process control, but.
You know, even if you don't want to
go that far, where at least gives
you those suggestions where the
operator's like, okay, I can do one
of these three things and here's
the potential outcomes I can choose.
Right?
Versus I'm the least experienced operator.
I've only been running this
process for six months.
I've never seen this particular scenario,
and now I've gotta call my supervisor
who is at home with his cell phone on his
Denis: Yeah,
Jim: or they got the phone
in the other room and they
can't even get ahold of them.
So they make the.
They make a wrong decision.
Right?
I think those are the kind of places
where AI can actually be incredibly
Denis: mean, yeah, I use LMS
Jim: it's
Denis: That's what
they're built for, right?
Not for logic big decisions.
Jim: it doesn't seem like
that's where a lot of people
want to go, and I think it's an
Denis: Mm-hmm.
Jim: so if I'm a plant manager,
I'm looking for that and I've,
I've even, I've even counseled.
You know, not a necessarily
a plant manager.
It's actually a, a director level guy, but
he's kind of focused in this one plant.
I'm like, if you want an a win with AI
this year, here's something I think we
can do where you can claim that victory.
Do you have a spot?
We could do something like that,
Then you can claim that victory.
See, look, look at what we did,
and look at what the benefit
Denis: we're an AI company now.
Jim: Right?
Denis: Mm-hmm.
Jim: right.
You can make that claim.
Regardless of, yeah, it was a really
simple and, and, and quick win, you know,
but we found the appropriate place to use
this tool it's gonna save us money, right?
So therefore we wanna
do these other things.
Denis: Yeah.
Jim: That's what I, that's
what I counsel people to do.
And, and in so doing, when you do
stuff like that, you realize like,
well, how good is our documentation?
Is it even good enough?
Denis: Oh yeah.
That's a whole different pair of sleeve.
Right.
Isn't good enough.
Jim: right?
So, so do I But again, we're
back to this basic right?
Oh, I want to go build this
LLM that helps my operators.
Well, if your documentation sucks
Denis: Yeah.
If you
Jim: and
Denis: have garbage in, you
can always put garbage out.
Yeah.
Jim: correct, right?
Like somebody sent me their TCCs last
week, we've built this little TCC
solution both in data park and in in
pi, and they sent it to me and I'm like.
Okay, here's this scenario and
you're telling the operator to
do this particular operation.
Well, what does that even mean?
Denis: Mm-hmm.
Jim: How do they know they've done that?
Right?
What, what tags are involved?
Like what does that procedure,
what does that even look like?
Right?
Because it was like these two sentences.
I'm like, because have to be a super
experienced operator to even understand.
What you're telling me, and more
than likely, a really experienced
operator already knows what to do.
This doesn't help them at all.
And there's not nearly enough
information to help a young in a young
Denis: Yeah.
Jim: 'cause it doesn't tell
you what does that even mean?
And those are the kind of things that,
you know, I, I think so many companies
like don't realize is like, we don't have
the basic stuff to be able to build these
incredibly great solutions on top of.
And yet they want it fast now and
you know, they want it to work.
And that's why I think a
lot of 'em are failing.
Denis: Yeah,
Jim: So,
Denis: so the solution is to indeed
start small with something concrete.
Focus on the basics.
They're important and they still work.
And don't forget about your future,
which is the people and the culture.
Oh, yeah.
Jim: for sure.
And you know, and the reality
is like, like I told this
guy, I said, here's the thing.
You
Denis: Mm-hmm.
Jim: out really quickly to
every plant very quickly.
As long as you have the same level
of documentation, like that's
literally something we could do
small that would scale really well.
Denis: Yeah, makes sense.
Jim, this, this was
Jim: So
Denis: it was a great conversation.
Before I let you
Jim: yeah.
Thanks.
Denis: me a bit more about, let's
say people want to know more
about you or your company, where
can they find more information?
Jim: Yeah.
So couple places.
Um, we do.
I, I do a ton of posting on LinkedIn.
Denis: That's true.
Jim: two minute on Tuesday video
that sometimes is a six or seven
minute video if I'm really ranting.
Um, but you can, you can find
me on LinkedIn, just, you know,
search my name, Jim Gavigan.
Um, we have a LinkedIn, YouTube,
I mean a LinkedIn company page.
We have a YouTube channel.
Just look for industrial insight.
Uh, obviously our website,
industrial insight inc.com
um, is a good place to find some
information about us and some of my.
Um, pontifications over the years.
There's a lot of my old blogs, which
I probably ought to go back and reread
and refresh, know, after 6, 7, 8 years.
'cause I bet they still
Denis: Yeah.
Jim: different technology on top now.
Right.
But, but yeah, those are the common
places to, to go find us is, is
Denis: Yeah.
Jim: places.
I don't, I don't participate in, I mean,
we have a Facebook page, but it's the
same stuff you'll see on the other pages.
So.
Denis: We'll put the links below in
Jim: But
Denis: notes.
Jim: Sounds
Denis: All right, well, in that
sense, Jim, I wanna really thank
you a lot for this conversation.
Um, I really liked the
Jim: Absolutely.
Denis: something concrete at the end.
Jim: Yeah.
Denis: That's always good.
And yeah, put the links below and I'll
thank you to the dear listener, of course,
for giving us, uh, the time to listen
and yeah, I guess we'll see you in the
next episode of What's In Your Data.
Thanks, Jim.
Good for having you.
Jim: Yep.
Denis: All right, take care.
Jim: you.
Take care.