Design, Engineering, Production and Maintenance

Product Design

Graco is starting to review AI for product innovation by feeding customer feedback into product design. Data from technical service inquiries is run through an AI-based call summarization and sentiment analysis which will be transferred back to R&D teams to ensure product managers and engineers can review data for product quality. This allows the teams to incorporate real-life user questions and difficulties into the new product design process. “We all want visibility into what is happening with our products. When customers and distributors repeatedly call with questions about usage or troubleshooting, our engineers need to know. AI summarizes every call, allowing agents to focus on delivering value while providing engineers with a consistent, thorough summary of each inquiry. Our engineers and agents love it,” Doreen Giorgi, Director of Global Customer Service at Graco said.

At Marmon, the R&D team is using AI tools to streamline patent, academic, and market research for new product development. “We don’t just create a new product because somebody says they need another widget,” explained Tom Hewer, Vice President of R&D at Marmon Heavy Duty Transportation. “We go much more in-depth now by using AI tools to investigate patents, explore academic research, and analyze market conditions. We combine this with voice of customer data, and all these AI tools really help us accelerate that work,” Hewer said. 

Designing AI into the product itself is taking off as a growth area for any type of smart electronics device. Farooque Munshi, Americas Data and AI Leader for Manufacturing of EY, works with automotive parts suppliers and OEMs that are shifting from bolt-on AI features to built-in AI. These systems can be embedded into vehicles to enhance their intelligence, personalization, and safety through features such as AI-powered cockpits to replace traditional dashboards and proactive diagnostics for real-time alerts about potential failures. “It’s the beginning of a journey to use the AI-first lens for the entire product lifecycle,” Munshi said. 

Precisely this is happening at Marmon. Tom Hewer told us about plans to use AI to interpret audio signals provided by brake pad sensors. AI cuts through the road noise to analyze specific sounds coming from sensors. “We've constructed our own models in-house to interpret this data, and it is providing more in-depth information about the product than we thought would ever be possible.”

Production Design

PepsiCo is using AI to create photorealistic digital twins of production and warehousing. The digital twins can simulate plant operations and the end-to-end supply chain. Working with Siemens and NVIDIA, PepsiCo can “recreate every machine, conveyor, pallet route and operator path with physics-level accuracy, enabling AI agents to simulate, test and refine system changes - identifying up to 90 percent of potential issues before any physical modifications occur.” 

Athina Kanioura, Chief Strategy and Transformation Officer at PepsiCo, explained why the company chose a 50-year-old warehouse as a test case: “We wanted to start with one of our most challenging facilities, because if you are able to drive benefits and unlock KPIs – financial KPIs – in a warehouse that is as old as the Gatorade plant…just imagine the potential everywhere else in the world.” During the first three months of the project, PepsiCo realized a 20% increase in throughput and 10 to 15% reduction in CapEx by uncovering hidden capacity and validating investments in a virtual environment.

GALLO created an AI-driven MRO intelligence dashboard to scan overall MRO (Maintenance, Repair, Operations) inventory. Nitin Murali, Vice President of Supply Chain Excellence explained, “AI is able to crawl through our asset data and give us very specific recommendations. It can tell us, ‘Hey, this transformer has been sitting there without any movement for 12 months. Do you want to consider divesting this unit?’ If we want to divest, it can tell us what kind of value we release and how it connects back to our overall manufacturing waste elimination strategy.” Ghost asset removal offers numerous benefits including improved readiness for audits as well as increased accuracy for tax, insurance, safety, and compliance records. 

Throughput

For 67% of manufacturers, efficiency gains are a top AI priority. At one manufacturing company the ultimate goal is optimizing capacity and maximizing throughput. The company is using AI to determine the right amount of production at each plant. One senior executive explained: “We are optimizing the speed of machines, finding out what would be the sweet spot for this machine to run without breaking anything. AI is helping us identify the potential of a machine – taking age into account – to understand what our actual capacity is.”

In the chemical industry, LSB Industries is targeting a similar goal with a focus on reliability and process optimization. Scott Bemis, Executive Vice President of Manufacturing explained, “One part involves getting that early notification, a sense that something has changed or is not going in the right direction, even before things like vibration start to appear. The other part is looking for additional capacity that we haven't been able to tap into in the past. We’re doing this by running our processes a little closer to shutdown limits.” In the future, Bemis’s vision is to have a dashboard showing the health of each of his plant sites blending information about the equipment as well as ambient temperatures enabling him to “take the pulse of the organization each morning.”

Physical AI for autonomous and semi-autonomous shipbuilding is in early testing at military shipbuilder HII. The company signed agreements with GrayMatter and Path Robotics in early 2026 for autonomous welding and metal surface finishing robots to increase throughput. According to Path Robotics, its physical AI model for welding “transforms a traditional industrial robot arm from a rigid, repeat-only machine into a real-time perception and decision-making system that can see, understand, and adapt to the variations of a shipbuilding environment.” 

Maintenance 

LSB is using AI for asset health management to receive notifications before alarms or automatic shutdowns. Scott Bemis explained, “Let's say the temperature in a vessel typically runs at 100°F, but it's now hovering at 101°F. AI will flag that and suggest ways to investigate. That would happen before the system sounds a high temperature alarm at 105°F or shuts down at 110°F. As a result, we receive an earlier indication about the health of the asset which allows us to start troubleshooting sooner.”

Snowflake’s software is using AI to reduce diagnostic time for a global industrial automation leader’s variable frequency drives. “These machines are very complicated and require a specific skillset to maintain and repair,” Clay Richard, Solutions Engineer at Snowflake told us. "If a specialist is needed, repairs can take hours. With our solution, the company was able to use AI to reduce diagnostic time by 65% from 32 minutes to 11 minutes,” Richard said. Minutes can be valuable if the drive breakdown is stopping an entire production line.   

Charter Manufacturing is using AI for a predictive maintenance solution that “combines vibration and sensor data from the plant floor with traditional machine learning and deep learning plus generative AI,” Jared Noble, Director of Digital Technology explained. “Gen AI enables natural language communication with the agent to look at predictive maintenance on our shop floor.” 

This approach helps Charter avoid unplanned downtime while also supporting asset life cycle management. “If we look at motor sensor data, vibration data, you can anticipate motor failure by vibration changes long before the motor fails. And rather than needing to replace a motor, we might be able to perform maintenance that keeps the machine running for a while longer,” Noble said. Whether an asset needs replacement or maintenance, the key metric is avoiding unplanned downtime. 

Charter’s pilot program saw success in the initial phase of the pilot program. “The system identified motor vibration data that prevented downtime that would have cost us significantly. We already saw the pilot program paying dividends before the true scale out,” Noble said. The next step for Charter is AI analysis of years of historical maintenance and machine performance data to provide recommendations for action in advance.

Latent AI is helping companies use AI to analyze edge data from their machines to reduce latency by bringing the AI closer to point of data creation. “Machines are operating 24/7 and there are sensors collecting information about those machines,” Natalia Jurado, Embedded Software Engineer told us. “Instead of leaving that information on the production floor, we can capture it and use it to take action.” A great example is energy usage from a machine. “The data may show there is suddenly a different energy profile which means something is going on. This is very powerful information, and that’s what we want to take advantage of, especially since it never needs to leave the factory floor,” Jurado concluded.  

Using AI for predictive maintenance offers benefits that go far beyond improved uptime. As Saar Yoskovitz, Co-Founder and Executive Chairman at Augury, points out, “The top metrics that people track are uptime and unplanned downtime. But if I can go from schedule-based maintenance to condition-based maintenance, maybe I don't need to fix the machine as often. Right now, one out of three spare parts is replaced for no good reason. That’s 33% waste, not to mention labor costs and safety. Spare parts are sitting in warehouses collecting dust and aging out of their warranties. Machine health is not a maintenance problem, it's a supply chain problem, a talent shortage problem, and a sustainability problem. It goes way beyond maintenance and impacts business goals and viability. AI is allowing us to rewire the whole ecosystem to be based on AI insights in real time.”

Belden is using AI for supervisor shift notes across its operations. “At the end of each shift, our supervisors recount what has happened and document anything of note that has occurred. These logs are reviewed by a lot of people, but it is difficult to identify trends or know the context of some comments. Now AI interprets all those shift notes and provides a concise summary at the end of the month,” Logan Cooper, Senior Director of Digitalization said. The natural language processing capabilities of AI enable it to categorize and tag information that can be difficult to search and analyze manually. The tool can also identify recurring themes in notes that may point to a pending machine failure.  

Cross-Functional Case: AI for End-to-End Correlations

One communications equipment manufacturer is using AI for a pilot program that takes a holistic approach connecting multiple phases of operations – from production to environment to quality – and then linking that data to each individual product by serial number. 

“We’re using data from our processes including details about individual machines (parameters, operator, maintenance history, etc.),” according to a senior leader at the company. “We also include data from the environment – temperature, humidity, barometric pressure – and connect that with in-process tests on a sampling of our products as well as final tests on all our products. Environmental data can affect our products, so we put all that data together to develop correlations.” 

The company uses AI to write algorithms based on the data it has collected. “It's a complex problem and there are multiple relationships between the product, the environment, and the testing. We know the correlations are there, but there are so many different factors. There’s not just one little knob to turn,” one senior leader said.

The goal is to predict what will happen before things go wrong. “The key is being able to pull the right information and correlate that across a particular line for a particular product and ultimately to the individual serial number,” the manufacturing leader explained.