AI Will Fail Manufacturers in 2025

Summary

If you are an average manufacturer in the U.S., 2025 is not the year to invest in artificial intelligence (AI). If your goals are to modernize and improve your manufacturing process, AI is simply not a practical solution. In fact, you should likely avoid AI like the plague to maximize return on investment (ROI). Any investment in AI will lead to remorse. This seemingly anti-AI message is, however, not an anti-progress message.

In fact, the reason companies should avoid AI in 2025 when they look to improve their manufacturing is because there is a near guarantee that you are better off investing in tools that supplement your factory’s human intelligence (HI) instead. I place my automation honor and an American dollar bill on this bet.
 

The state of American Manufacturing

There are around 250,000 manufacturing companies in the U.S. While the data is not entirely conclusive, it’s estimated that somewhere between 50 percent and 70 percent of these manufacturers have no form of automation deployed in their manufacturing. It’s 2025; we are focused on reshoring manufacturing. Yet at least half of manufacturing in this country is driven by our stagnation and a shrinking human labor pool. That’s a problem.

We’re talking Industry 4.0 while the majority of manufacturing lives in a sub-Industry 3.0 world (Figure 1).

It’s more efficient to invest in HI

Your control engineers and operators know your machines and processes. Many can service and diagnose issues in real time by sight, sound and occasionally taste. These people are valuable resources. Their experience is an under-leveraged HI resource. These are the people who need to be empowered with access to time-series control information first.

Business systems can benefit from information locked in operational technology (OT) networks, but that value will pale in comparison to the value your operations teams can add. This is the team of individuals who can logically comprehend the relationships among system variables. They know what additional vibration and heat lead to. They see an encoder skip and know how a slowing actuator will impact the process. Giving these teams historical records of data from all inputs on a machine will vastly increase their ability to do root-cause analysis. The quicker they solve problems, the more efficient your operations become.
 

Bridging IT and OT

Connecting people or information technology (IT) systems to OT networks is the first step needed to empower either AI or HI (it’s also one of the more challenging steps). OT is the hardware and software that monitors and controls devices, processes and infrastructure used in industrial settings. This is all the technology we use and deploy on the factory floor and includes industrial automation communication protocols.

The main purpose of OT is to turn the physical world into digital signals; basically, to connect and represent real-word variables as data. IT, on the other hand, must take data and turn it into information for users and other systems to ingest.

For IT to get real value from OT data, that data must be modeled. This gives each data point the context, meta data, state and values needed to make it usable in IT systems. Data context is imperative for us to get the full value from our automation expenditures. To leverage data analytics and AI tools, factory floor data must be given context. A PLC doesn’t need meta data; it is programmed to react to data values. However, that same data value moved to the cloud needs context. The context of this information will empower your operators and managers.
 

Operators are not going away

AI systems today could be trained on all of the information that operators have. Models can be created that can improve root cause analysis and support preventive maintenance, but today, this comes at a significant cost.

Training AI tools on your data and building these efficient models will keep a highly compensated data scientist, working with expensive tools, employed for a long time. This data scientist is great, but their system knowledge will come from your operations team, and unlike your operations team, this data scientist is not employed to fix anything. Trusting your operations team to leverage historic data with HI is a significantly more efficient place to start.

Focusing on HI empowerment is not a counter to future AI

Empowering your controls team with time-series data from your automated machines means you are moving forward on a digitization journey. This data access offers the best return by first being used by your HI. In the future, this same data access can be leveraged by AI. The smart manufacturer will allow AI tools to further mature, and first focus on the ample low-hanging fruit that HI can harvest.

_{This feature originally appeared in the May 2025 edition of Automation.com Monthly.}