The fifth edition of the “Real-Life AI” series looks at common mistakes manufacturers make when deploying the technology, from investing in unnecessary tools to overlooking the importance of human involvement. See what six industrial and technology leaders say companies should reconsider before putting AI to work.
What is one thing manufacturers are getting wrong when deploying AI?
Responses have been edited for length and clarity.
One common mistake is not taking a look at what your current technology vendors already offer. Before diving into new AI investments, manufacturers should conduct a digital audit to understand where the real gaps are, what foundations are already in place and whether they have the right people to operationalize AI. In most cases, the fastest adoption and strongest results come from using AI that’s already embedded in platforms your teams know and trust, rather than introducing entirely new tools that create friction and slow momentum.
Jacob Sanchez, Industry Solutions and Community Development,igus
Stop trying to force a circle into a square hole. If a certain type of AI is hitting your feeds but it's not good for your business and team, then pivot away. Don’t waste time, energy, resources and trying out AI that your team doesn’t even believe in or is passionate about. And realize not everyone in your organization has to be an AI champion and that’s okay.
Whenever you’re deploying AI, you also need to think about agentic process reengineering. Let’s take the order-to-receipt workflow. We know the outcome that we want. We want all of these lines that we’ve placed with our suppliers to arrive on time and, if there are any changes, respond to that. There is a way to achieve that outcome where AI does so much of the heavy lifting and the humans manage by exceptions now. You let the AI do most of the heavy lifting and you have to change the way that you work so that you do less work, but you coordinate with AI.
Trying to be too quick or doing nothing. If you ignore AI deployment, you will lose out. But if you just let every employee start using tools and bots, then it becomes very dangerous for large global companies. Start small, get adoption, see your data is capable and show successes. I also think that you have to be extremely vigilant on just deploying tools: the data ownership, the exposure to the public, combining GDPR and data tools. You cannot just install a "co-work" tool in a global company and not be surprised if your data leaves your building.
Thinking that people can be completely replaced by AI. We can have sensors for continuous monitoring, but eventually there is going to be a human that has to go fix the stuff, touch the materials and feel it. The smell of oil in a machine shop is not something a robot can learn. Completely depending on a fully connected system that can take its own inputs and its own decisions is a faulted model. I go back to my own research at MIT when I was a student. The most stolen item in the world, at least at that time, was Gillette razors. I was working on a research project where we were trying to put RFID chipsets on every piece that was sold in a Walmart store. But the economics did not work when you have to do this to millions of units. So similarly, when you think about sensors and connected AI, there is an ROI thing that needs to be calculated for every instance. You can’t just use a sensor, build an AI model and just forget about it.
Offloading too much thought to LLMs. AI is a fantastic sounding board for developing ideas, strategies or even running down shop analytics. However, giving too much credence to AI's outputs and drawing conclusions from them is something we all need to be mindful of, regardless of industry. AI has use cases, but trying to shoehorn in AI in every facet of your operation can create more issues than it solves.
If you would like to share how AI has impacted your business, contact Nolan Beilstein at [email protected].