The 5 Levels of AI for Manufacturing with Industry 4.0's Jeff Winter

 If there’s one word that can summarize the state of the manufacturing industry today, it would probably be “change”. 

Supply chain challenges, labor shortages, economic unpredictability, and a rapidly evolving global landscape have spurred manufacturing business leaders to rethink how they do business, and how technology can help them evolve.

The power and potential of Generative AI in solving real-world business problems is impossible to ignore, but what exactly does that look like for manufacturers?

Join Advisory Host Dave Horstein and Industry 4.0 Expert Jeff Winter as they break this down for leaders solving for business in a very specific way for the manufacturing industry.

global.hitachi-solutions.com

Transcript

Exchanges Podcast by Hitachi Solutions

Season 3, Episode 19

Airdate: Wednesday, September 20

Title: The 5 Levels of AI for Manufacturing with Industry 4.0’s Jeff Winter

Host: Dave Horstein

Guest: Jeff Winter

Dave Horstein
If there's one word that can summarize the state of the manufacturing industry today, it would probably be change, supply chain challenges, labor shortages, economic unpredictability, and a rapidly evolving global landscape have spurred manufacturing business leaders to rethink how they do business and how technology can help them evolve the power and potential of generative AI in solving real-world business problems is impossible to ignore.

What exactly does that look like? For manufacturers? Hello everyone, this is Dave Horstein and welcome to Exchanges with Hitachi Solutions. I'm excited to be able to answer this very question with Mr. Industry 4.0 himself, Jeff Winter. With that intro alone is not nearly enough to acknowledge everything Jeff does in the industry.

He's not just an Industry, 4.0 thought leader Jeff was recently ranked by Analytica as the number one thought leader in the world in this space, and has amassed over 80,000 followers on LinkedIn. On top of that, he's won 13 industry awards for being the top in the field. He's also known as being very prolific in the industry. Not only is Jeff very prolific in the industry, but he's known for being able to break down complex topics into very simple concepts.

Welcome Jeff. So happy to have you on today.

Jeff Winter
Thanks for having me here. I'm excited to do this.

Dave Horstein
Absolutely. The first question I have for you in kicking off today's discussion, so you and I've been talking about this a little bit since we met in April, but AI is it's on the cover of every magazine. It's the title of every blog post, but really as we zoom in today just to start, how should manufacturers really be looking at AI?

Jeff Winter
It's a great question because when I talk to industry leaders, I like to make sure that we're all understanding it the right way. And so I often emphasize that artificial intelligence should really be thought of as just a tool, a tool that is designed to replicate, and some cases improve upon human intelligence.

And when you think of it like that, there are two primary ways that you can use AI.

You can use it to help augment decisions or automate decisions. So let's break that down a little bit.

So the first is augmenting decisions. Imagine you're a plant manager and every day you're swamped with data from machine performance metrics to supply chain logistics data. Now, what if you had a tool that could sift through this data in real time, highlighting patterns and correlations and insights that might take days or even weeks to discern and figure out?

That's where AI comes in. It doesn't just replace your decision making.Instead, it enhances it.

It's like having a super smart assistant whispering invaluable insights into your ear, helping you make more informed decisions. That is augmenting decisions. The other is automating decisions, and so to talk about this one, let's shift gears a little bit.

 While augmenting is about enhancing human capability, automation is about letting AI take the wheel in specific scenarios. Think about quality control. We've all seen those assembly lines where workers are meticulously inspecting products. Now you can introduce AI powered visual systems into the mix and it can spot minute details, perhaps even those that are invisible to the human eye.

And it's not about replacing humans in this scenario. It's about reallocating human resources to tasks that require more creativity and intuition and strategic thinking.

In fact, there was actually a study done in 2020 by MIT in collaboration with BCG that breaks down AI into five different levels, or modes of integration into decision making. Level one was human generates an AI evaluates level two was AI generates the insights and the human uses it in decision processes.

Level 3 is the AI recommends, but still the human decides Level 4 is the AI decides and then the human implements and level 5 is the AI decides and the AI implements so full automation. What's interesting about this study is that the companies they're seeing the most significant financial benefits are those that are using all 5 levels, not just level 5 where the AI decides and AI implements.

So my take on this study is that those that are really using AI right, they know that the goal isn't just about blindly handing over control of decisions to AI. It's about understanding where AI can be most effective and where human intervention remains paramount. It's a delicate balance, ensuring efficiency without compromising on the unique human touch.

Dave Horstein
I like that is especially in in the type of work that we tend to do. A lot of the conversations I've been having with leaders from other organizations is around human involvement with AI and the good news from my perspective is a lot of leaders. It's they're not asking me how they can replace head count with AI.

They're asking me how we can improve and optimize their business, leveraging humans to provide context to AI, and I think trying to follow the levels that you came, that might be like a level 2 Level 3 on the scale that you provided.

But I think that's good news to a lot of individuals that maybe are wondering like is what is the impact to my job going to be in the coming years?

But what?

What I'm basically responding with is OK, yes, we can focus on, you know, how humans are going to provide the context to the artificial intelligence. That's going to mean it from an organizational development perspective is you're going to have to rescale or upskill your employees. And so that's where it becomes a really interesting dynamic where we're thinking about optimizing the business, using automation and official intelligence.

But it requires that human component into your point, hitting all five of those, those levels within the within the model where humans still have that critical part of it. That's encouraging. Well, let me, Jeff, playing off of that response.

Then you're having a lot of these conversations with leaders, C-Suite boards at the in, in manufacturing. How are manufacturers leveraging AI today?

Where are they seeing the most value in it?

0:7:39.570 --> 0:7:43.540
 Jeff Winter
So another good question, and luckily there's some data to kind of back this up. So the manufacturing Leadership Council is part of their manufacturing. In 2030, Project released their 2023 study on the Future of Industrial AI and manufacturing.

 Just a couple months ago, and while there's a global competition for AI dominance, manufacturers still remain cautious, with only 57% experimenting to identify the best application and only 29% have incorporated AI into formal corporate plans.

 And only 28% are implementing AI operationally. However, on the positive note, 96%, so almost everyone are expecting AI investment levels to rise and rise pretty dramatically. So if you look at kind of where it's being done, of those who are doing it, the top corporate function that has begun to adopt AI manufacturing is in the actual function of manufacturing and production.

 And that honestly, shouldn't be a surprise, as the core function of any manufacturers to transform raw materials into finished products for consumption or use.

 Therefore, manufacturing data arguably becomes the most important and valuable data within the organization to drive improvements to drive change, and to optimize processes.

 Now if you dive deeper into the specific application areas for manufacturing production, it was plant floor IO T analysis coming into first place at 40% then preventative maintenance coming in at second at 36% process improvements at 30% and quality improvements at 30% also.

 And in case you're curious, it's just the the overall manufacturing industry.

 The second area of adoption was inventory management and the third was actual R&D and this actually really aligns with research that was done by IoT Analytics and their industrial AI report as well because they actually showed predictive maintenance came in as the first new case.

 Quality inspection was number 2 and manufacturing process optimization was third and supply chain was fourth.

 But supply chain actually had the highest growth rate out of all of them.

 And if you look at another study that's out there, the World Economic Forum, in their 2022 White paper called unlocking value from artificial intelligence in manufacturing, they do a nice breakdown about the potential of AI and where it could provide the most value.

 The two studies I just referenced talk about where it's being adopted and this one talks about kind of where it can and should be used to gain the most value and they break it up into six major areas.

 Production process with examples including the process optimization, line balancing, product design.

 The second was quality.

 The third was maintenance, and that's both in the planning and the predictive maintenance.

 4th is health and safety, including employee health and safety by incident prevention, as well as process safety.

 The 5th is supply chains, most notably demand and pricing forecast, and also warranty and service management and the six was energy management.

But one interesting thing from that manufacturing Leadership Council report I just mentioned is that how manufacturers are measuring success of their AI initiatives.

Unfortunately, just 22% said that they had a specific set of metrics in place to measure AI deployment effectiveness and impact, and 61% said that they had no metrics in place. That's crazy. I mean that tells you that quite a few companies and manufacturers specifically are rushing into AI without thinking through all the details on kind of how they're going to know if it works or not.

But what's clear here is there's a huge opportunity for manufacturers to better demonstrate how AI is making an impact.

0:11:33.420 --> 0:11:46.840
 Dave Horstein
That, selfishly is incredibly validating for me because I've a lot of the conversations I've been getting pulled into are, like everyone wants to talk, use cases like I probably have two or three meetings a day where it's just, oh, we just want to see what can this do for our business?

 What can this do for our business and I can I can see because we're doing a lot of video lately. I can see the reaction when I first ask how are you planning to measure the success of this engagement? And it's usually a puzzled look. Like, what do you mean in my recommendation to all these leaders is to make? Don't think of AI as just a pie in the sky.

 As you said, it's a tool. It's really just a tool. Don't think of it as an end all be all it's going to solve these problems at the board is putting pressure on on you to fix.

 Start with your strategy.

 How do you measure the health of your Business Today? And let's figure out how we can extend AI so that it can help you there.

So many organizations that I'm observing are going about it the opposite way, and they're chasing use cases, which I think is a good way to kind of dip your toe in the water and get a feel for the potential.

 But there there's to me a whole a huge opportunity cost of doing so by not tying back to the strategy that your team has spent likely years cultivating.

 So I appreciate that insight, Jeff, and we'll see if it holds true. But I think that it's so important it AI or any other change that leaders do have some sort of metrics that they're evaluating the effectiveness again.

 So it's really surprising that only 22% have those metrics in place today. Very interesting hitting on that same topic, Jeff, what would you say is, do you have an example like a real world example of AI truly making an impact on a manufacturing business?

0:13:21.190 --> 0:13:34.740
 Jeff Winter
So I like to go over this one cause before I just went over kind of the the aggregate statistics, but the use cases are more meant to be aspirational and to help aspire or sorry inspire companies to show them what they can do.

 But this doesn't necessarily mean it will work for your organization this exact way, but I do have a couple different examples that I can share that I really like and one of my favorite use cases for AI manufacturing is actually Microsoft. Most people forget or don't realize they are a manufacturer, not just a software company.

 They actually have over 42,000 SKUs at 33 manufacturing facilities and distribution centers and their manufacturing team in China and the US integrated AI to enhance their operations based on a strategy with three main objectives get connected, become predictive and grow to be cognitive, which means train the system to ideate solutions that enhance human abilities.

 So the system that they deployed not only monitored individual machines, but also provided an overview of all machines simultaneously, and this holistic view allowed for early detection of potential issues preventing machine failures or production halts and data was used not just from inventory, but also even from the sales and their promotion campaigns that they had.

 Factory workers, though initially expressed concerns about the the real time data being shared with executives. Kind of fearing premature reprimands, but with everyone acting on the same data, they saw significant improvements and it changed the mindset of everyone kind of working on the factory floor.

 So here's some examples of the improvements that were made:

 The order commitments improved from 40% within five days to 95% within 48 hours and a $50 million year over year.

 Reduction in errors is what they saw, and they saw savings of 10 million from waste reduction in process optimization, 15% more accurate demand forecasts and one of the standout outcomes of this transformation was a more engaged and informed workforce across all levels.

 But there's actually a ton of examples out there.

 Two more that I'll kind of share with you.

 Another example is Nissan, so they're running AI predictive maintenance platform to do their Arul or their remaining useful life predictions 1 / 7500 assets, and the company claims an unplanned downtime reduction of 50% and a payback period of less than three months.

 And Nissan scaled the solution from 20 critical assets to thousands without increasing the workload of the on site preventive maintenance team, which is incredible.

 And another example is Micron Technologies, who I just heard them speak at the recent critical manufacturing conference that I was at a few weeks ago and their entire presentation was on AI manufacturing and they're using machine learning algorithms to detect anomalies earlier and more precisely than humans, especially when it comes to quality issues like scratches and holes that are hard to detect.

 And they even use AI with acoustic listening to identify unusual noises in the manufacturing process, because an hour of downtime for them cost around $250,000. But their results were also impressive.

 A 10% increase in manufacturing output, a 35% reduction in quality issues, 25% faster time to yield maturity and ultimately saving them millions of dollars through early detection of issues.

0:17:5.650 --> 0:17:9.520
 Dave Horstein
OK, you're turning me into a believer, Jeff, I'll, I'll give you that. So for a long time it did a lot of this felt kind of pie in the sky and and as I mentioned, dipping the toe in the water and trying to figure this AI thing out.

 Those are some real, tangible results. Those Microsoft numbers are pretty impressive, but Nissan as well? Umm this it's it is eye opening to realize the potential impact here and tying back to your earlier point about human involvement, it's nice.

 I'm trying to remember now which was the reference around more transparency.

 That was Microsoft, right? Greater transparency to the workforce.

0:17:43.810 --> 0:17:43.950
 Jeff Winter
Yes.

0:17:44.650 --> 0:17:45.980
 Dave Horstein
Yeah, that. So that's something we've been preaching. I've been doing.

 I had been in the CRM space for quite some time and I was always advocating for an increase in visibility of data.

 Let people see the health of the business and how they impact it, and it's going to drive their behavior.

 And so if AI has the potential to open that up, and an even larger scale and faster, that is going to, that's going to consistently yield a more empowered and engaged workforce.

 So like I said, you're turning me into a believer.

 I'm getting on the train with that said, so manufacturers that are are looking to get into this space and and better understand AI.

 What advice would you give them? If they're really looking at ways that they can make an impact with their AI initiatives.

0:18:31.470 --> 0:18:40.860
 Jeff Winter
So when I think about manufacturers diving into the world of AI, I'm reminded of an age old adage quality over quantity.

 It's not about implementing the most advanced AI tools available, but rather about addressing the unique challenges and opportunities that your manufacturing processes face.

 And there's two ways to look at this and I got this concept from the book autonomous transformation by Brian Evergreen.

 So it's problem solving versus future solving.

 So everyone talks about the problem solving side or solving problems and using technology to help solve those problems.

 Identify your biggest pain points, your challenges, and see where AI can make the biggest impact and have the quickest wins.

 And if you do that, you will see results with this and some great examples of questions based off of potential problems include things like where are routine decisions being made all the time, where if people were freed up, this could save a lot of time and money.

 Or where are their decisions being made that are quite variable or inconsistent or we are decisions being made where the result could have a profound impact and probably more than anything.

 Where are people relying on their gut instead of data?

 And this is a great and easy way to get started, but that will rarely result in a company transformation.

 To quote Brian Evergreen in his book, “The Problem with problem Solving is that solving a problem is inherently directed at that which you do not want, not at what you do want.”

 And I love that quote because it makes you realize problem solving is a very myopic view of the whole business, the entire world that we live with in.

 And this is where future solving comes into play.

 What do you want your future to look like?

 How should things be?

 What should you be doing and where?

 And this opens up your thinking to be much broader and allows for a new way of thinking that can result in truly transformative initiatives that take advantage of artificial intelligence.

 So that's a big piece of advice I would give is to start thinking about future solving.

 The other advice I would give is if you really want to make an impact with AI within your business, you need to switch away from just reporting on results to influencing outcomes.

 So these are, you know, if you think about it, those that are thriving with AI have migrated away from the mindset of, hey, look where we're using AI and all the different places that we're using it and instead they're focusing more on look at how we've made an impact with artificial intelligence.

0:21:20.40 --> 0:21:23.270
 Dave Horstein
I this this all what's great about this Jeff?

 Is it all tied together? So, we talked a few minutes ago about starting with strategy and having measures for success.

 And I think that goes really closely along with moving away from reporting results in really thinking more about, you know, future solving versus problem solving.

 And so it sounds to me like what we're really looking at is is making sure we're we're having that forward looking view as we're planning an AI engagement or even a strategy around the use of it.

 But as we're approaching that and starting and talk about integrating AI into the business, what does that path look like?

 So how are we recommending organizations and specifically manufacturers begin integrating AI into their business?

 Is it some sort of like a maturity road map or a pathway that manufacturers can align to?

0:22:14.180 --> 0:22:15.320
 Jeff Winter
Yeah, absolutely.

 I mean, there's actually quite a few different AI maturity models and road maps that are out there.

 I mean, if you just Google it, you'll come up with with dozens, you know, integrating AI into business, especially in manufacturing, though I would say if you look across all of them, they typically follow a similar path.

 So first is all about awareness.

 You need to understand what AI can do and what it can't do for your specific business needs. I mean, a lot of people think that it can do way more than it actually can.

 And that's a panacea for solving all problems.

 That's not the case.

 Maybe one day it will get there, but right now it's really good at specific things and you need to understand what it's good at and what it's not.

 You know, generative AI is a great example of a whole bunch of people misapplying the technology right now because they think it can do more than it really can.

 The second is exploration.

 Start with pilot projects, perhaps focusing on areas like predictive maintenance or quality control where there's a higher adoption rate.

 And there's more.

 Tried and true, you know, tested methods out there that you can learn from.

 The third is integration.

 Once you've seen success, begin integrating AI tools within your core processes, ensuring that your team is trained to work alongside those tools.

 4th I would say is optimization.

 You need to continuously refine your AI models based on real world feedback and performance, because one thing that makes AI unique is once again, it's trying to replicate and improve human intelligence, which means that as it starts to get more and more data and produces more and more outputs based off of utilizing artificial intelligence, it can have a feedback loop to continually make itself better based off of what it's already learned.

 And finally, as transformation at this stage, AI becomes a strategic driver, not just a tool.

 Reshaping business models and strategies.

 Because remember, the journey is as much about people and process as it is about the technology and collaboration.

 Continuous learning and staying updated with AI advancements are absolutely key in this journey.

0:24:28.790 --> 0:24:36.560
 Dave Horstein
Thank you so much, Jeff, that that makes a lot of sense to me and it's very consistent with the way we've been providing guidance to our own clients.

 I can't Thank you. Enough.

 We have hit everything from the concept of artificial intelligence in manufacturing to some very specific use cases to how you can measure success to even how to implement AI in a manufacturing organization.

 We've been kind of all over this wild ride of artificial intelligence.

 I wrote down four things, so usually I do three, but there were four really good Nuggets that I took away from this conversation, but I think our listeners should really bring home with them the first point you made that really resonated with me is that we should think of AI as really just a tool.

 And you broke it down and it to really 2 segments.

 One is it either augments decisions or two.

 It automates decisions and if you think about that from a use case perspective, it can help ground you in real practical potential use cases within an organization and not just pie in the sky.

You know we can solve the world's problems with AI.

 The second thing that you hit on was manufacturers are not holistically evaluating the effectiveness of their AI implementations.

 Only 22% of manufacturers surveyed in this study you referenced had reported that they came up with KPIs to measure the health and effectiveness of their AI engagements.

That is a problem that needs to be resolved.

 Leaders need to be thinking of how they're going to measure the effectiveness of their use cases, especially so that they can continue to prioritize investments in this area.

 The third thing I took away was really we should prioritize quality over quantity.

 You were getting into the idea of problem solving versus future solving.

 Where problem solving is thinking about getting rid of the issues that we have today, rather than harnessing the power of this technology and really thinking about where you want to go as an organization in, in figuring out ways AI can help you get there.

 And finally, the fourth major takeaway for me Jeff, is moving away from reporting results and shifting towards influencing outcomes.

 That is so impactful because as we know, we've talked about on this podcast multiple times.

 When we become data driven or data enabled as an organization, we're moving away from being reactive to proactive and AI should be a tool that helps you get there. Being proactive and becoming the organization you're really seeking to become, Jeff.

Thank you so much. I know you're kind of in the middle of I think it's a multi continental speaking tour and virtual engagements and manufacturing exhibitions and all of that.

 So thank you for carving out time and talking with us today.

 Our listeners, thank you so much for hanging out with us and hearing from Mr Industry, 4.0 himself, Jeff Winter and if you'd like to learn more about Hitachi Solutions, check us out at global.hitachisolutions.com and we'll talk to you next time.

###