The first installment of the "Real-Life AI" series asked manufacturing and technology leaders, "Where has AI actually delivered ROI." In this second edition, those same sources revealed where companies are wasting the most money on AI and why some investments never deliver meaningful results.
Their answers offer guidance for avoiding common AI investment mistakes.
Where are companies wasting money on AI?
Responses have been edited for length and clarity.
Arturo Buzzalino, SVP of Product & Chief Innovation Officer, Epicor
Companies waste money when they treat AI as a science experiment with standalone pilots, generic chat tools or even custom models that are disconnected from their goals and value statements. If AI isn’t anchored in core processes and clean operational data, it can easily create more complexity than value. Even if you do have great data and are AI-ready, that readiness does not equal execution. You can have strong data foundations and still lack the organizational habits, workflows and discipline required to make AI useful at scale.
Microsoft Copilot is a really good example. Every company has two options: You either buy something horizontal or you buy something vertical. If you think your AI transformation is going to be delivered by a horizontal platform; it’s never going to happen. You can have hundreds, if not thousands, of uses on Microsoft Copilot and see absolutely zero gains in productivity. The reason that happens is because it’s not specialized into a specific workflow. What you get with these horizontal tools is they’re cheap, they’re easy to build simple things on but they don’t complete entire workflows. To achieve that 80%-90% time savings for specific roles in the manufacturing plant, you need something very verticalized, very specific that specializes in that workflow.
In my opinion, its customer service chatbots—especially in technical industrial environments. High-quality customers can always tell that a chatbot is answering. Every chatbot will try to please you. I have not seen a chatbot that, after three questions, did not disappoint or frustrate me. If you have enough high-quality data and technical expertise that an LLM or other tool can really utilize, maybe. But it's very difficult to give good chat answers. Another area where I see money wasted is buying AI tools without good, structured data and data governance. It's so hard to get good results when the available data is not clear or requires human interpretation that people historically have in their head.
Jacob Sanchez, Industry Solutions and Community Development, igus
They are wasting money in two ways: by not using it at all, and using it way too much. If you aren’t using it even a little bit, you are walking right into a sink hole. Companies spending tens or hundreds of thousands of dollars on corporate subscriptions and credits is insane. We are still working through environmental impacts of AI usage, and one thing that is not good is when companies spend all this money on ‘trials.’ Truly successful companies set up their AI workflows so things are perfect or near to it within one to four attempts, not 10-plus.
There are two places I see. One is, there is a sense of using AI for AI's sake. The second one is, big tech companies are saying employees are going to be measured on their performance based on AI tokens used. We use tools like Claude to see some of the data collection exercise—the coordination, reaching out to people and collecting the information which is already out there. That process we try to accelerate. But I, or anybody on the team, cannot go in front of the board and say, ‘My AI model said we should invest in this product, or this is the product we should launch.’ The accountability eventually sits with the human being.
If you would like to share how AI has impacted your business, contact Nolan Beilstein at [email protected]
