Survey: What 500 Manufacturers Want You to Know

Ed Ballina
Well, hello team. I’m Ed Ballina.

Alvaro Cuba
Hello guys, Alvaro Cuba here.

Ed Ballina
Lots going on, right? Especially FIFA. But Alvaro is the master and he’ll talk about that. Welcome to the Manufacturing Meet Up podcast. This is a show where we kick back on our downtime and we have real chats about what actually happens on the floor and address some of the more common issues that supply chain and ops people are facing today. Alvaro?

Alvaro Cuba
What’s OPEX?

Ed Ballina
Well, thank you. Yes, the hat. So I am doing some shameless self-promotion today. That is my company, Operational Excellence Consulting. Thank you. Three and a half years and still alive and kicking. Sometimes I wonder, but no, it’s all good. So I thought I’d wear that today. I haven’t self-promoted shamelessly in a while, so this was it. And you have a really fancy, looks like a golf or a globe.

Alvaro Cuba
No, this is my tennis one. We just finished Roland Garros and I was so happy to see so many young people. The women’s winner, a teenager, 19 years old, Andreeva, which is amazing. And in the semi-finals, three of the four men’s semi-finalists are guys between 18 and 21. I’m very excited that young people are coming and it’s going to bring a lot of excitement to tennis. Amazing talent. What’s for today, Ed?

Ed Ballina
Amazing talent, right? Rejuvenates the sport. Beautiful.

Today we have an interesting conversation about the State of Production Health. As you know, Augury has partnered with IndustryWeek for the last several years to put together a State of Production Health report. How’s it going? What are the trends we’re seeing? Since 2023, by the way. And the 2026 report just dropped. And let me tell you, it is everything that Alvaro and I had discussed at the end of the year.

Looking into our non-existent crystal ball. But the speed at which it’s happening, our crystal ball has been shattered. Crystal ball is blown. So I won’t steal the thunder. We got some interesting stuff to share with you. Alvaro?

Alvaro Cuba
We are going to talk about the biggest findings, and more importantly, what it means to you in the plants or in the reliability programs. We also found some interesting contradictions that we think are the core of the study that we’ll be discussing, and we’ll end up as always with some concrete opportunities that you can tackle and use going ahead. But before going into that, please subscribe, click the button at the bottom of the screen so you don’t miss these conversations. And with that, let’s get started.

Ed Ballina
Let’s do it. Hit the button.

All right, so hopefully we prepped you enough that you are anxiously awaiting these findings. Well, they are pretty interesting. So in 2026, the two biggest issues that survey respondents pointed out were workforce constraints at 43% of respondents and unplanned downtime at 40%. Those are the two top manufacturing challenges. Downtime is up 10 points from last year alone. Folks, that’s not a little degradation. That is a step change. And to be honest, we suffered a lot from COVID, right? A lot of setbacks, a lot of issues, supply chain, couldn’t get parts. We expected to recover from that, and we kind of have not gotten back to pre-COVID levels in most places. And now to get this information, that’s really surprising. But I will tell you that they’re related.

Why do you think our unplanned downtime is up? Maybe it’s because of issue number one: workforce constraints. If you don’t have the right people, if you don’t have the right experience base, or even the numbers, your machine doesn’t run as well. And it’s something that Alvaro and I have been poking at for quite a while because of the loss of experience we’re seeing in the field. And that can certainly explain it.

Quality, yield, and throughput issues score number three at 37%, and capacity constraints at 30%. Now folks, put the whole formula together. They all make sense, right? If your downtime is up 10%, every time a line goes up and down, you increase the risk of making off-quality product and you lose yields on startups and changeovers. The more you do that, the more it impacts those numbers.

And of course, if your lines are down 10% more, then you’re going to have capacity constraints. You’re either going to have to source product in or you’re going to have out-of-stocks with your customers. Bottom line: manufacturers are struggling to get productive capacity out of their plants. And it’s not getting any better, and the people running them are really challenged.

Supply chain issues are at 22%. That’s a generic term for parts we can’t get on time, raw materials, et cetera. And it’s nearly doubling from last year. So we’ve got work to do. Alvaro.

Alvaro Cuba
These challenges aren’t new. The high cost of materials, energy, quality, yield, and throughput issues have been consistent since 2023. And probably the one that is accelerating is workforce constraints and upskilling, which is jumping in the list in 2026.

But if you think about the figures that Ed was showing, what is happening is the pressure is moving from outside. In previous years we were talking about logistics constraints and material shortages. Now we are talking about internal things. This is a good thing because we have control over those things.

And I think the people who answered the survey are in alignment, because when the survey asked what they want AI to help with, it’s consistent with that. They are saying: improve quality, yield, and throughput and reduce unplanned downtime. OEE is also ranked the most important metric at 45%, which is in sync with that.

The only thing, and this is a cue to go to segment two, is: why, if that is happening, is the pressure continuing to mount? I think it’s because before we didn’t have the tools, so we didn’t have the results. Now we’re starting to have tools and AI, but the results are not consistent yet. That mounts the pressure. We’ll be discussing in segment two a little bit more about AI and how we can use these tools to convert into results that really help you in your day-to-day operation in the plant.

Ed Ballina
Well put, Alvaro. I scratch my head just like you. If you look further in the report, you’re going to see that we’re spending a ton of money on AI to address these same issues. So we’re spending a lot of money, but we’re not seeing the return. Why? I have two opinions. One is our environment is changing faster than we are adopting AI. And number two, some of our solutions were so widespread and systemic in nature that they haven’t yielded results yet. There’s a lot of infrastructure that has to go in for all these massive learnings to work.

So that brings me to segment two: the AI numbers are really, really big. And how big are they, Ed? They are big.

Alvaro Cuba
Very optimistic.

Ed Ballina
Yes. If you take nothing away from this report, take away this: we have issues, we understand our issues, but we have technology to help address them. Here are a couple of key highlights. At the top of the list of AI implementations is predictive maintenance. We’ve moved from breakdown to preventive to now predictive. And the application of AI is really yielding very solid returns in that area. The use cases are at 57%, up 22 points from last year. That’s not an evolutionary improvement. That is a revolutionary improvement. Think Niagara Falls. A transformation.

And what that points to is something we’ve been bringing up on the podcast: the first place where you would want to start in an operations environment is machine health, because it underscores and underpins everything about your line. But interestingly, number two is process quality, yields, and all that. That’s fully to be expected: you start with machine health, you look at the opportunities in process health, and they start coming together.

What was interesting is that the opportunity on the process side was called out to be even higher than machine health. Even though we have a plethora of machine health suppliers — you know which ones we root for — process health has remained a bit of an open space. There are lots of people going in there, but we think that’s still a bit of blank space right now.

A lot of scale up happening: 42% have now scaled more than half of their AI pilots. So it’s starting to go from piloting to large-scale implementations. And when you have a machine that fails, you’re taking product out of the consumer’s hands, causing yield issues and waste. And anytime your machine goes down, you increase the chance of a human-machine interaction, which carries a probability of causing an injury. They all work together. Alvaro.

Alvaro Cuba
As Ed was saying, AI is maturing, and more importantly, you in every plant — it doesn’t matter the industry or the size of the plant — are noticing that. A couple of figures: 83% are planning to spend more on AI in 2026. And the other thing that shows commitment and also a more realistic picture is this: before, “very advanced” in AI implementations was at 41%. Now it has reduced to 30%, which means you are getting a grasp of reality and becoming more grounded about what you think you can do.

The focus is shifting from experimenting to scaling. 42% have reached scale in 50% or more of their pilots. And one important thing, which I think is the biggest qualifier here, is that the focus is changing from doing things separately here and there to a more connected communication layer. What does that mean?

AI predicts the failure. Then the algorithm recommends the fix. Then the generative AI writes the work instructions. And then it sends communication to the operators who implement the instructions. If you put together the different pieces and go from data and insights into execution, that’s where the real value is. And that’s what different plants are just starting to understand.

More areas are now quantifiable, which is good. People are now able to quantify benefits from uptime, from yield, from quality, and that’s improving. The conclusion is: maturity is going up. The understanding of the end-to-end is starting to percolate into the plants and generate results. And this is in line with what we discussed in our previous episode about a McKinsey study saying exactly the same thing: AI can predict and provide the right knowledge to workers in the right moment. That’s the game.

Overconfidence is going down, which is good. If you are overconfident, you are not really paying attention to the opportunities, and you are only putting more and more pressure on your people. We were talking in segment one about why implementations are not connecting to results. It’s because we need to connect the different dots all the way from insights into execution.

Ed Ballina
This is like building a house. One of the messages you’re starting to pick up from us is that agentic AI is winning — a consistent message we’ve been relaying for quite a while now.

As we move into segment three: we threw a lot of data and numbers at you. What’s the big takeaway? Bottom line, as Alvaro has said, there’s a lot of confidence in AI, but there are also some roadblocks. Some of it, we’re learning things we didn’t know we didn’t know. So we start scratching our heads and saying, wow, there’s some work here.

So the report says manufacturers are struggling with workforce issues, downtime, capacity, and some supply chain challenges as well. Part two says they’re confident in AI and doubling down, putting a lot of money into it. But part three says even with that, there are still lots of roadblocks. From a simple standpoint, the math doesn’t tie. We’re having problems, we’re spending a ton of money on AI — you would think that would mean we’re overcoming some of them. But I think we’re finding things we were not aware of.

Poor data quality is at 47%, up 20 points from last year. And that did not even show up in the top questions a couple of years ago. Cybersecurity was at the top for three straight years. That has now come down. Operational hurdles at 43%. We’re starting to recognize that our data quality is suspect. These models — garbage in, garbage out — can only intake what is provided to them. And if it’s not of good quality, that’s a problem. This is where the whole data ontology discussion starts to come up. But that’s food for another day.

Ed Ballina
The other piece that really struck a chord was workforce obstacles. 41% of respondents said they need to upskill or improve the skill of their people, and 38% said they are dealing with labor shortages. It’s all about people. And it’s part of the conversation that Alvaro and I have been having: you can have the best technology in the world and invest all the right funds, but if the people side — meaning skill and capacity — isn’t available, then it’s just a lot of fancy paperwork and fancy charts on a screen. People, people.

That being said, the bottom line is the study is telling us we have challenges. But these are roadblocks in a maturation process, and that connection is really part of the message.

Alvaro Cuba
A couple of good things about the roadblocks that I think are interesting: leadership buy-in is only 22%. It has come way down, which means that in 80% of cases, you now have the backing of leadership. That’s not surprising — they are hearing all the news about AI and they want to cash in on it. And the other thing that went down is bringing people along with you, the support and buy-in from the workforce. It went down from 50% to 30%. So that’s good.

But I think the most important thing, as Ed said, is that the biggest roadblocks are data quality and upskilling your workforce. And if you think about it, those are exactly the easiest ones to address with AI.

In segment two, we talked about how positive people are about AI and how much money is being invested. If you start by putting part of that effort and money into data quality — let the sensors get the data, let the sensors send it to AI, and let AI do the analytics, the analysis, and the flagging of issues. And as Ed was talking about with generative AI, let AI do the instructions, the manuals, and the next steps. That’s relatively easy and minimum cost. You don’t even need to pay for the sensors now. Suppliers put the sensors on the machines, do all the analytics, and send the communication. They charge you as you go, once a month.

The second one is upskilling. Agentic AI can now do the work instructions, the maintenance procedures, the troubleshooting guides. It can generate, contextualize, and build actionable things that go straight to your people in the moment they need it. Think about it — you yourself have a ton of questions for any topic. Now you go into ChatGPT and ask a simple question. You get the answer you want: short, straight, and it asks, “Did I answer your question? I can also tell you A, B, or C.” So you don’t need to study A, B, and C. You only need to study exactly what you need in that moment, which reduces the need for large upskilling processes.

And you can use AI to attract people. The new generation is drawn to AI. They love those kinds of tools. If you explain that the work they will be doing is interacting with machines and using AI for day-to-day work, they are going to love it. So in essence, the two biggest challenges are the two easiest to solve. Cost effective and very fast. Implementation of quality data can take two or three weeks. Nothing more.

Ed Ballina
It’s that fast. All of us are using ChatGPT or Claude or some similar tool. If you are a plant manager listening to this, what are the first few moves we think you should take? First, start with machine health. You’re building a house here — build a strong foundation in your equipment and your lines. Machine health is number one. But based on what we’ve just learned, start looking at your data quality. I would have said this before I read this report. Your issues with data quality are going to hold you back eventually. Don’t wait for that to happen. Start taking action now to make sure your data is clean.

Alvaro Cuba
A couple from my side. The first is what we said: AI and technology are starting to have clarity of purpose. Think about it — investments are going up, the pilots are successful, you are now able to quantify the impacts, you have the support of your leadership, and your people are interested. So it’s becoming the perfect moment. And the second, the biggest thing, is to close the loop between the machine and the person. AI and high technology are not coming to replace people. We need more people, not less. But those people, we need them prepared.

Ed Ballina
Amen.

Alvaro Cuba
Let the technology do what it does better: get the data, get the insights, get the knowledge you need, and then make sure it connects well with the person who has to make the final decision and who has to act on it. Closing the loop between machine and operator is the key to getting the benefits of this transformation. With that, I don’t know if you wanted to say something else before we start the closing.

Ed Ballina
It’s the well-worn term we use: until you close the last yard with a human being, you get nothing in return for all these huge investments.

Alvaro Cuba
And that brings your people with you as well. So friends, this is the wrap-up for today’s episode. Thank you very much for joining the Manufacturing Meet Up. If you enjoyed the episode, please follow or subscribe. Like this episode if you are watching us on YouTube, or leave us a review if you are listening on iTunes. And please share this with your pals and help us grow the meetup.

Ed Ballina
Our message doesn’t change: we want more of you, and hopefully you want more of us. The way you do that is email us at mmu@augury.com. We also have a link to the report we talked about — you don’t have to go searching for it, it’s in the show notes. And I think by the time this comes to you it might be somewhere around the Fourth of July. So in case I don’t get to say it, happy 250th anniversary to the United States of America.

Alvaro Cuba Thank you guys, bye.

Ed Ballina Thank you.