Jeff Witt says the worst case for a mistake in enterprise generative AI analyzing sales figures is redoing the work.
“if I'm using generative AI or an agent to analyze some sales figures, worst possible case is that I have to redo it.”
Building the Foundations for Scalable Industrial AI - with Jeff Witt of GE Vernova
15 Oct 2026 · 20m · Jeff Witt · 20 quoted facts
In brief
Jeff Witt of GE Vernova argues that AI on a factory floor must be built and judged differently from office AI. A wrong answer there can damage equipment or endanger a worker, so an accuracy level that is fine for sales analysis is unacceptable. Trust comes from recommendations that reflect the specific machine and its history, not generic chatbot answers.
The central tension is between starting AI quickly and scaling it. Many manufacturers launch one use case and stall. Witt's answer is slower groundwork: automate only processes that are already documented, optimized and shared across sites, and invest early in a consistent data structure. That upfront cost makes every later use case faster, and worker feedback keeps the system improving.
Key insights
Factory AI is judged by safety, not accuracy
Witt says enterprise teams may accept 80% accuracy, but a forklift system cannot miss a pedestrian 20% of the time. The cost of an error includes damaged equipment and injured people, not just redone work.
Edge computing replaces the cloud for control systems
Some shop-floor equipment is deliberately air gapped (cut off from outside networks) to block malicious access. Models therefore run locally near the equipment rather than making round trips to the cloud.
Generic answers destroy frontline trust immediately
Frontline leaders already know the machine, its history and its operators, so a Google-style answer fails the first test. Witt says you get one shot at a first impression.
Human review creates a built-in training loop
When people evaluate, plan and carry out an AI recommendation, that process becomes feedback for the model. A simple thumbs up or thumbs down shows which suggestions are useful and which are noise.
Earn the right to automate before scaling
A process should be documented, optimized and lean before AI touches it, and ideally be used across all 35 sites. A consistent data structure built once, with a standard template per asset type, speeds every later use case.
Questions
Why can't companies use the same AI approach in a factory as in an office?
Witt says the consequences are different: a bad office result means redoing work, while a bad factory recommendation could wreck equipment or put a technician next to a running machine. That raises the standard of accuracy and care required.
What does it mean to run AI at the edge?
It means running the model on computers close to the equipment instead of sending data to the cloud and back. Witt says this gives more control over security, which matters because some factory systems are intentionally isolated.
Why do frontline workers distrust AI tools?
The host cited a study saying only about a quarter of frontline workers are excited about AI. Witt explains that workers already have deep context on their machines, so a generic recommendation looks useless and loses credibility on first use.
Why do most factory AI projects never get past one use case?
The host cited a figure that only 12% of manufacturers move beyond a single use case. Witt says companies often skip the groundwork, such as documented, optimized processes and organized data, that makes scaling possible.
What does having good data foundations mean on a factory floor?
Witt describes a context layer: a clear asset hierarchy (division, plant, line, work cell, pump) and a unified name space, meaning one standard template so every pump's data looks the same. It costs effort upfront but speeds up every later AI project.
Facts
Stated by the speakers. Each quote matches the transcript, which shows what was said, not that it is true.
Jeff Witt says 80% accuracy, which may be acceptable in enterprise AI, is inadequate in operational technology settings such as forklift pedestrian safety.
“I can't have a forklift truck not hit a pedestrian. 80% of the time. That's not good enough”
Jeff Witt says some levels of equipment on GE Vernova's shop floors are intentionally air gapped.
“some of our, uh, levels of equipment are air gapped intentionally, right?”
Jeff Witt says data from shop-floor control loops is not sent to the cloud to run models against it.
“We are not going to round trip something to the cloud to analyze or run a model against that sort of data.”
Jeff Witt says running models on local edge computing gives GE Vernova more control over security than cloud processing.
“we run edge compute locally to to bring that model down, you know, a little bit closer to the edge, where we have more control over the security.”
The host cited a PwC and Manufacturing Institute study finding that only about a quarter of frontline workers feel excited about AI.
“the PWC and manufacturing Institute study from the spring found that only about a quarter of frontline workers come into AI feeling excited about it.”
The host cited a study finding that over half of leaders lack confidence that their frontline teams are ready to lead through AI adoption.
“over half of leaders aren't confident that their frontline teams are ready to lead through this change and adoption.”
Jeff Witt says frontline leaders hold deep context about their machine, work cell, people and history, which sets the bar for AI recommendations.
“they know the machine, they know the work cell, they know the people that are working on that machine, they know the history”
Jeff Witt says GE Vernova has up to ten years of history for much of its equipment, part structured data and part unstructured notes.
“for a lot of our equipment, we might have history going back ten years”
Jeff Witt says a human reviews, plans, schedules and resources each AI recommendation, and that execution becomes feedback to the model.
“a human in the loop is going to evaluate that recommendation. They're going to plan the work, they're going to schedule the work, they're going to resource it”
Jeff Witt says shop-floor workers lack time to stop and query a chatbot, so chatbot interfaces do not work in operational technology environments.
“Do not have time to stop, go to a terminal, pull up our phone, start asking questions, right?”
The host cited a figure that AI runs on the floor at at least 42 percent of manufacturers, but only 12 percent have moved beyond a single use case.
“AI is running somewhere on the floor, at least at 42 of all manufacturers. But only 12% have gotten it past a single use case.”
Jeff Witt says his team's principle of earning the right to automate requires that a process first be fully documented and optimized.
“we have to earn the right to automate. The way we are embracing that is. Has the process been fully documented? Has the process been optimized?”
Jeff Witt says GE Vernova scouts for mature, well-documented, efficient processes used across all thirty-five sites as AI candidates.
“those processes that are mature, well documented, efficient processes that are used in all thirty five sites.”
Jeff Witt says the value of industrial AI lies in eliminating waste across all thirty-five sites running the same product, not in success at one site.
“The value's not in being successful in one site, it's in eliminating that waste across all thirty five sites that are running that same product.”
Jeff Witt says GE Vernova is investing in a context layer including a unified name space so that all data flows through one template.
“building a unified name space so that all data flows through this one template.”
Jeff Witt describes an asset hierarchy running from division to plant, line, work cell and pump, designed to be understandable to the newest employee.
“this division, this plant, this line, this work cell, this pump, it makes sense, and they can follow it”
Jeff Witt says the upfront work of structuring and contextualizing data pays off across later AI use cases, from the third to the hundredth.
“that basic structure and context work pays dividends. When you look at the third use case, the fiftieth use case, the hundredth use case”
The announcer states that Jeff Witt's views on the program are his own and do not reflect those of GE Vernova or its leadership.
“the views and opinions expressed by Jeff on today's program are his own and do not reflect those of GE Vernova or its leadership.”
Jeff Witt says he first looks for an internal standard, then to his network or the wider industry, before building a new solution.
“I look internally first to see if we have that standard. If we don't have it internally, then I look to my network or to the industry”
Chapters
- 0:02:54Why industrial AI is held to a higher standard
Witt explains that safety consequences change what counts as good enough, making factory AI far more demanding than office AI.
“you might expect 80% accuracy and be perfectly happy with that”
- 0:04:55Edge computing and air-gapped equipment
Security and control-loop constraints mean models must run locally on the shop floor rather than in the cloud.
“Bringing the model down to the edge and running it locally”
- 0:06:57Context is what earns frontline trust
Frontline leaders know their machines deeply, so AI that offers generic advice fails at first contact.
“You have one shot with the AI recommendations to make that first impression.”
- 0:10:11Feedback loops that keep improving the model
Human review of recommendations, captured with simple thumbs up or down, feeds back into the model.
“as simple as a thumbs up or a thumbs down interface on the recommendation”
- 0:10:59Why the chatbot interface fails on the shop floor
Workers have no time to type questions, so AI should use their location and task as context.
“on the shop floor, that interface does not work”
- 0:12:50Earning the right to automate
Witt's team scales AI only on mature, documented, lean processes shared across many sites.
“we have to earn the right to automate”
- 0:14:45Building data foundations that scale
A unified name space and asset hierarchy cost effort upfront but accelerate every later use case.
“building a unified name space so that all data flows through this one template”
- 0:18:39Final advice: use existing standards
Witt closes by urging teams to follow existing standards and treat AI as a new tool, not a new process.
“It's a new tool, but it's not a new process”
Transcript
0:00:12
Host
Welcome everyone to the Emerj AI in Business podcast. Today's guest is Jeff Witt, Global Process Digital Authority at GE Vernova. Jeff discusses why industrial AI can't be measured the same way enterprise AI is, since the margin for error on a factory floor involves physical safety, not just redone work. He explains how context determines whether frontline workers actually trust and adopt an AI tool, and lays out a practical framework for proving a process is ready to automate before scaling it across multiple sites.
0:00:41
Today's episode is sponsored by Poka. A quick note for our audience that the views and opinions expressed by Jeff on today's program are his own and do not reflect those of GE Vernova or its leadership. In this episode, we cover what separates industrial AI from enterprise AI. To go deeper on this topic and learn how to evaluate AI vendors by assessing leadership expertise and why funding benchmarks can signal product maturity and stability, download our free PDF report.
0:01:07
Five ways to select the right AI vendor at Emerj dot com slash A I V one. That's E M E R J dot com slash A I V one to download your copy. Now, the conversation with Jeff. Jeff, welcome back to the Emerj AI in Business studio.
0:01:41
Jeff Witt
Thank you for having me. I'm glad to be here.
0:01:43
Host
Great conversation waiting for us. So if our listeners remember correctly, the last time we had you on the show, we spoke about computer vision. And this time, a very interesting conversation about how we should be thinking about AI in the industrial environment. So when I have conversations about AI outside of this environment, it feels as though most people think that AI is AI, regardless of which direction you're pointing it. So that's where I want to start today.
0:02:09
Is what is genuinely different about applying AI in an industrial environment versus an office one?
0:02:15
Jeff Witt
I think there's two components to this that make it significantly different. So first would be consequences. So if I'm using generative AI or an agent to analyze some sales figures, worst possible case is that I have to redo it. I have to correct the data, I have to redo some work. If I'm using AI on the shop floor to make recommendations, it potentially could tear down a piece of equipment. If I do something incorrectly, worse than that, I could maybe put a technician right next to an operating piece of equipment, a spindle that that's still running.
0:02:54
So there's a safety component, there's a equipment safety impact component that just doesn't exist in in the enterprise AI environment. So it is a little bit different. It changes what good looks like, right? So for enterprise AI, you might expect 80% accuracy and be perfectly happy with that. In the OT space, I can't have a forklift truck not hit a pedestrian. 80% of the time. That's not good enough, right? So the way it's judged is completely different.
0:03:28
So that just adds a whole different layer of rigor to an AI project on the shop floor.
0:03:34
Host
Absolutely. So what's the one thing that a software team building AI capabilities or AI programs for officers always gets wrong when they try to apply it to a factory? Is there a general assumption?
0:03:46
Jeff Witt
I think there's constraints on the architecture that we have in the OT space that are definitely not present in the SAS subscription space. So I can securely access data in my cloud environments. All I need is the proper SSO login credentials and I can get access to that data to run a a model or to analyze the data. There's a a rather complex architecture on the shop floor that some of our, uh, levels of equipment are air gapped intentionally, right?
0:04:22
So that people can't, uh, maliciously get into our equipment. So running models in the cloud against systems on the shop floor that might be a control loop that are actually controlling the equipment or controlling a process. We are not going to round trip something to the cloud to analyze or run a model against that sort of data. So just the architecture itself has to be different in the OT space than it is in the enterprise space.
0:04:55
Bringing the model down to the edge and running it locally is usually the solve for that. So instead of round tripping to the cloud, we run edge compute locally to to bring that model down, you know, a little bit closer to the edge, where we have more control over the security.
0:05:14
Host
Hm, oh, that makes a lot of sense. And I think it's basically the same as building a house for domestic use versus building a big commercial building. It's the same, both AI, but you have to treat it differently, both houses or both buildings. But the safety requirements will also be completely different. So, Jeff, this is the part where I try to get more practical and I'm going to to read from my notes here. Cause a very interesting concept that jumped out to me was that the PWC and manufacturing Institute study from the spring found that only about a quarter of frontline workers come into AI feeling excited about it.
0:05:49
Most of them are very skeptical about it, and over half of leaders aren't confident that their frontline teams are ready to lead through this change and adoption. So this isn't really a technology question anymore. It's a what does this tool need to do for the person on the floor, for them to trust it and not question it anymore? And given the work that you've done all over in asset health and predictive maintenance, what AI capabilities would meaningfully support the unique workflow environments in industrial settings?
0:06:18
Jeff Witt
I agree with that statistic. What I've observed is that when a frontline leader is making a decision today, they have a lot of context. They know the machine, they know the work cell, they know the people that are working on that machine, they know the history, they know what's happened over the last couple of weeks, and they know what's planned, what's upcoming for that machine. So there's just this huge ecosystem of context that a frontline leader knows inherently just by working in the environment.
0:06:57
That is the expectation, right? So if the AI comes in and it gives a really generic recommendation, it goes out and searches the web and it recommends, Hey, when I see this signal, I suggest you generically perform this function. You have one shot with the AI recommendations to make that first impression. And I think what the mistake people make is they start with this general feedback, general chatbot type feedback loop, and it's it's not gonna work.
0:07:34
You have to start with that very specific operating context of that pump, that machine, those operators, this product that we're running on that machine. And we have all of that data, but if if you, if you keep it superficial, they're not going to see the value in the tool. And so my personal opinion, and I've seen it, is you have one shot at this. And if you put an AI recommendation engine or a coach or something out there and it gives a super generic Google type response to your first question, you've lost credibility, right?
0:08:11
It has to be very, very specific.
0:08:13
Host
I can see that, but now also at the same time, the problem of data. So you just say, we, we do have the data. There's no reason for it to have to go and give us generic answers. But how do we capture that knowledge? How do we capture the knowledge of someone who's run a line for over 20 years without turning it into a major documentation project?
0:08:33
Jeff Witt
There's two things, right? So for a lot of our equipment, we might have history going back ten years. So there is a lot of context. Some of it's structured data, a lot of it is unstructured notes, documentation, things like that, but that the AI can access that quickly. But the important thing to know about putting a new workflow on the shop floor that's AI enabled is that we have this built-in feedback loop. So right now, if the AI makes a recommendation, a human in the loop is going to evaluate that recommendation.
0:09:08
They're going to plan the work, they're going to schedule the work, they're going to resource it, they're going to make sure we have the right tools, the right parts, the right documentation. It's a safe environment, they're going to do everything they normally do, but the recommendation came from the AI. But all of that execution part is now a feedback loop to the model, so it can now take that feedback and make itself even better. That's a component of feedback to the AI model, the training of that model, that is very, very valuable in in improving it over time.
0:09:43
So if we can go into it and it can give a lot of context, and it knows the history of the asset, and then we augment it with this feedback loop to improve it over time. Now you've got something.
0:09:54
Host
So basically, we're using existing data to start everything up, give it enough context to start, and then continuously just develop it with up-to-date data as we go. And it becomes this super capability driven machine. Is that the the right idea, the right way to think about it?
0:10:11
Jeff Witt
Yeah, it. It's as simple as a thumbs up or a thumbs down interface on the recommendation, right? I really like that interface because it's I give the AI recommendation a thumbs up. I liked this, this was good, this was valuable. Show me more recommendations like this, or a thumbs down. This was not helpful to me. Please don't show me more like this. This is just noise in the system, right? But having having all of the context and then having this built-in feedback loop to the to the model so that it can improve over time.
0:10:44
Very valuable, especially as you start to scale them out.
0:10:46
Host
Yeah, I like that. Is there a a capability that everyone assumes is essential that you would think, uh, ranks pretty low in how essential it is? Something that you think, oh, you know, we can actually live without this. It's nice to have, but it's not worth the hype.
0:10:59
Jeff Witt
So where AI started gaining traction was the chatbot, and honestly, on the shop floor, that interface does not work. Do not have time to stop, go to a terminal, pull up our phone, start asking questions, right? It needs to know already that I'm standing in this work cell in front of this machine and give me some context and provide those recommendations to facilitate my work, to make me more efficient, more safe, more productive. And use that context as an interface rather than just an empty box waiting for a question.
0:11:43
That interface does not work on the OT environment.
0:11:47
Host
Yeah, that's interesting, cause I think you're probably the third person this month to say that, to say, you know, we, we really passed the era of chatbots. And to be quite frank, it it never was that useful. Is that the? And it sounds like you're saying exactly the same, so definitely out of that era and not using that anymore. So that's interesting. I like to think about this environment as something, and and AI in general, this year specifically, it's not something that new anymore.
0:12:15
Everyone is trying to adopt some kind of AI. Um, and this is where I have a a nice little stat again. So what we're seeing is that the number going around right now is that AI is running somewhere on the floor, at least at 42 of all manufacturers. But only 12% have gotten it past a single use case. So it's not that the companies are stuck at point zero, but they're getting stuck at point one, so they've started and then that's it.
0:12:42
Um, so for leaders in the industrial environments that are at that threshold of transformation, what practical steps should they consider first?
0:12:50
Jeff Witt
I would agree with that statistic. It is difficult to scale, and I think it's because we are not starting in the right frame of mind. So I've heard the team here say something recently that's kind of stuck with me, and it's we have to earn the right to automate. The way we are embracing that is. Has the process been fully documented? Has the process been optimized? Have we applied lean principles to this and taken all of the waste out of this process already?
0:13:25
If yes, then great. This is a good candidate of a use case that we could probably augment with AI, automate some portion of it, and then potentially scale, right? So that's what we're scouting for is those processes that are mature, well documented, efficient processes that are used in all thirty five sites. And so, though that would be an ideal candidate that we could then take and apply AI tools to that, elevate that process, automate it, and um, begin to scale that out to multiple sites.
0:14:09
Because you're right, we don't want to do a single use case, no matter how successful. The value's not in being successful in one site, it's in eliminating that waste across all thirty five sites that are running that same product.
0:14:24
Host
And that's exactly something that I also wanted to ask you today. Is, if we're thinking about good documentation, optimization, everything that you basically mentioned, what does that actually look like? I think it's something that we hear the phrases. Or, you know, we have to make sure that we have good data foundations. But what does good data foundations really mean on a factory floor? In plain terms, how do we know it's good enough to start with?
0:14:45
Jeff Witt
We all know we have lots of data. What we generally are lacking is the context and the information model that lays on top of the data. So we do believe that that is very important. We're investing in it. It doesn't really exist today. It it hasn't been something that we've needed, but to facilitate and accelerate AI projects, we do see the value in building out our context layer, adding context to all of our data sources, building a unified name space so that all data flows through this one template.
0:15:30
And and building an asset hierarchy that makes sense to even the newest of employees. So if they walk in and they get access to our data set, they can follow the structure. Okay, this division, this plant, this line, this work cell, this pump, it makes sense, and they can follow it. And then once they get down to that layer of that pump, it has a pump unified name space template laid on top of it. So no matter what pump you pull up in the organization, you're gonna have the same data structured the same way.
0:16:10
And that's that's an investment, right? It's an investment to do that, uh, and build that out, especially for the first couple of use cases. You might look at it and say, well, we don't have the time to do all of that. But that basic structure and context work pays dividends. When you look at the third use case, the fiftieth use case, the hundredth use case, because it's a build it once and then just maintain it, improve it, and continue to use it forever.
0:16:38
So every AI use case after you begin this organization of the data just is accelerated and much quicker to implement.
0:16:47
Host
Yeah, I like that because it's it is a scary concept, thinking about, uh, what makes the data ready and what makes for a good foundation. But if you break it down to that, it, it becomes more practical and becomes, uh, more doable. Um, it, it takes the scary part out of it. Jeff, it's. It's been a great conversation. I'm looking at my notes here and I don't even know where to start. I think for me, the biggest takeaway is so starting with getting the difference between industrial AI and enterprise AI, and something that I think will really stick with the audience as well is that you said the biggest difference is consequence.
0:17:17
Um, with industrial AI, it's kind of a life or death situation, which is not necessarily more important or more serious, but the human is involved and the human safety is involved. That just it, it changes the equation a little bit. Um, and when we're thinking about the AI capabilities that will be meaningful in this area, context from the worker is important. And not relying on generic information, because, like we mentioned, generic information is not gonna build trust.
0:17:43
In fact, it's going to just that. I think that's where we see AI pilots stall, or just something being stuck in pilot mode. Because no one is really using, and it's not gonna scale because no one trusts it. Because it's just something that they can also get on Google, nothing special on there. And starting in the right state of mind, I like that, I like thinking about it. And more than just, oh, this sounds like a good plan, let's make sure that our plan has a plan.
0:18:07
And and that's where we ended off today. Anything else that you would like to leave the audience with?
0:18:12
Jeff Witt
I, I, I'm a big believer in standards and followings, existing standards. And I look internally first to see if we have that standard. If we don't have it internally, then I look to my network or to the industry and see if that problem has already been solved and documented somewhere out there. And then if, and then we take those standards and we build with the people doing the work, right? And that's how we get to success.
0:18:39
So don't reinvent the wheel. It's a new tool, but it's not a new process. We're just eliminating waste in the in the processes. We're just trying to be more efficient.
0:18:50
Host
Oh, what a great final thought. Jeff, thank you so much. thank you for your time. thank you for coming back. I know that we have more to talk about, so we'll definitely get you back again for a third time. But it's been wonderful having this conversation with you.
0:19:01
Jeff Witt
Thank you so much. Happy to be here.
0:19:17
Host
Wrapping up today's episode, I think there are three key takeaways from our conversation with Jeff. First, the standard for good enough AI changes completely once a recommendation can affect physical safety, not just redone work. Second, context, not generic, searched up answers, is what earns an AI tool trust from frontline workers, often within the very first interaction. And finally, scaling AI across many sites starts with earning the right to automate, documenting, optimizing, and improving a single process before expanding it further.
0:19:49
In this episode, we covered what separates industrial AI from enterprise AI. To go deeper on this topic and learn how to evaluate AI vendors by assessing leadership expertise and why funding benchmarks can signal product maturity and stability, download our free PDF report. Five ways to select the right AI vendor at Emerj dot com slash A I V one. That's E M E R J dot com slash A I V one to download your copy. For further executive level analysis and to join our network of leaders delivering workflow impact with AI, visit Emerj dot com.
0:20:21
On behalf of the team at Emerj, we'll see you on the next episode.