AI in Manufacturing: A Practical Guide to Safe and Strategic Implementation

Is artificial intelligence a threat to your workforce or the key to unlocking unprecedented efficiency? During a recent Moore on Manufacturing episode, Mike Sibley and Kevin Golden hosted three James Moore experts, Julie Kniseley (HR Services Leader), Tomas Sjostrom (Technology Services Leader) and Daniel Shorstein (Digital Team Leader), to discuss the realities of AI in manufacturing. This conversation revealed that successful AI adoption isn’t about choosing between innovation and caution. Iit’s about implementing both simultaneously through clear communication, robust security measures and strategic planning.

Understanding AI Beyond the Buzzwords

Before manufacturers can effectively implement AI in manufacturing environments, they need to move past the hype and understand what artificial intelligence actually means for their operations.

Daniel Shorstein offers a straightforward definition: “I like to define artificial intelligence as computers acting as humans and making decisions similar to how humans would make those decisions.” This distinguishes AI from traditional programming, where inputs generate predictable outputs. With AI, the system processes information more like a human analyst would, with outputs that can vary based on multiple factors.

Generative AI vs. Traditional Automation

The conversation highlighted an important distinction within AI technologies. Traditional automation follows rigid rules, while generative AI can produce creative and novel outputs. For manufacturing applications, this means AI can now handle tasks that previously required human judgment, from cleaning messy data to making cost comparisons across multiple vendors.

Daniel explained how the new O3 reasoning model from OpenAI represents a significant advancement: “It will not just provide you with the answer right away. It will actually plan and think ahead.” These models can access tools like web search and run calculations, making them particularly valuable for research-intensive tasks like parts cost analysis.

The Human Factor: Communication Prevents Fear

Julie Kniseley emphasized that the biggest challenge in AI adoption isn’t technical. It’s human. “I think it’s inevitable so whether we are prepared for it or not it’s coming and the question becomes how does a business owner prepare the workforce for it when it’s that it’s kind of that fear of the unknown right it’s very scary.”

Starting the Conversation Early

The key to overcoming workforce resistance is transparent, early communication. Julie stressed this point repeatedly: “You have to start the conversations with your workforce early you have to figure out how even you’re going to use AI.”

Many employees fear that AI will eliminate their jobs, but Julie reframes this concern: “It might change their job. Their job, their focus may be different, but it’s there as an enhancement, not a replacement.” She pointed out that AI can handle the frustrating, time-consuming parts of jobs, freeing employees to focus on work they actually enjoy.

Multi-Generational Training Strategies

Implementing AI in manufacturing requires acknowledging that different generations have different comfort levels with technology. Julie noted, “A 22-year-old kid might be absolutely fine with changing technology because they’ve been playing computer games since they were 13 versus somebody that’s kind of on the older end that maybe just had the experience or the need to.”

Her recommendation? “Training, training, training.” But not just one approach. Some employees prefer video tutorials, others need classroom instruction and some will simply experiment on their own. The flexibility to accommodate different learning styles makes adoption smoother across the entire workforce.

Security Cannot Be an Afterthought

While Julie focused on the people side, Tomas Sjostrom brought crucial attention to the security risks that come with AI adoption. His message was clear: embrace AI, but do it securely.

The Growing Threat Landscape

Tomas warned about how cybercriminals are using AI to create more sophisticated attacks: “We’ve seen just over the past month an increase in business email compromise just because they’re getting better and better and better at, you know, creating fake emails that look real, fake websites that look real.”

Perhaps most concerning is the rise of deepfakes: AI-generated voice and video content that can impersonate executives. “You also have, you know, the existence of deep fakes, voice and videos that make it sound like your big boss is calling and wanting you to transfer $200,000 to this other account.”

During his presentation, an attendee who worked at a bank received a real-time warning from management about exactly this type of threat, underscoring how immediate and present the danger has become.

Data Privilege and Access Rights

One of Tomas’s most important technical warnings concerned data access: “When you pull that data and make it accessible to a chatbot, you bypass all of those data privileges and access rights.” This means sensitive information that’s normally restricted to certain employees could become accessible to anyone using the AI system.

For manufacturers, this could mean protected design documents, proprietary processes or customer information becoming inappropriately accessible. The solution requires careful planning about which data sources connect to AI systems and ensuring proper access controls remain in place.

Privacy Policies and Third-Party Tools

Tomas advised manufacturers to thoroughly review the privacy policies of any AI tools they consider: “Just so you know that you retain the rights to your own data, just so you know that no one can jump the boundaries and get access to your data even though they shouldn’t have.”

He also recommended establishing clear guidelines about which AI tools employees can use. Without such policies, employees often create free accounts with random AI services and upload company data without understanding the implications.

Strategic Implementation: Starting Small and Smart

The experts agreed that manufacturers shouldn’t wait for perfect conditions before experimenting with AI, but they should be strategic about where to start.

The Risk-Based Approach

Daniel recommended thinking about AI adoption through a risk-based lens: “You really just look at, okay, for the things that take me or take my people a lot of time, take a lot of energy that we don’t like doing. Also look at if I brought a generative AI tool into the mix, how can I identify the ones or how can I put a human review in place that mitigates that risk the most?”

He suggested avoiding high-risk applications initially, like automated resume screening, which could introduce bias concerns. Instead, start with lower-risk applications like flagging unusual transactions for human review, adding an extra layer of oversight without replacing existing controls.

Treating AI Like a Junior Analyst

Daniel offered helpful guidance for setting appropriate expectations: “Treat it like you might treat a very book smart but maybe not, you know, highly experienced professional.” This means applying the same “trust but verify” approach you’d use with a new analyst or manager.

“Come back, review it, see how well it did, see where it made those mistakes, and then just kind of remember that because it’s going to be fairly consistent in what it’s good and what it’s not good at.”

Infrastructure Preparation

Mike Sibley raised an important practical consideration: maximizing AI’s value often requires having data properly organized. “Part of the power of that is having your data set up in a right way so it can access all the things that it can.”

Tomas confirmed this observation, noting that he and Daniel had recently discussed building an AI solution for their own IT service organization. The key was having someone who understood both the business needs and the technical capabilities to facilitate productive planning discussions.

Encourage Experimentation Within Guardrails

All three experts emphasized that leadership sets the tone for successful AI adoption. Tomas stated, “I firmly believe that the best approach an IT leader or a business leader for that matter can have towards AI is to enable the usage of that and encourage the usage of it but lead it so it’s becoming a secure and efficient implementation.”

Daniel added that encouraging personal use helps employees develop competency: “The best way to get good at using generative AI is to just use it. You know, if you spend a few hours a day on it, even an hour a day on it, on anything, you’re going to quickly get comfortable with what it’s good at, what it’s not good at.”

He even suggested a meta-approach to learning: asking the AI itself how to use AI more effectively. “I’ll show her how you can just pop it open and say, ‘Hey, I’m trying to do this with generative AI. Can you think of different ways that I can do it?'”

Prepare for What’s Next

The conversation made clear that AI in manufacturing isn’t a distant future concern. It’s a present reality requiring immediate attention. Julie summarized the urgency: “I think that in another six months, you know, everybody’s going to have to get on board in one way or the other.”

Manufacturers who take a thoughtful approach now, establishing policies, training employees, securing their data and experimenting strategically, will be better positioned than those who either rush in recklessly or wait too long to start.

The key is finding the balance that Tomas articulated: embrace AI and enable its use, but lead the implementation with security and strategic thinking at the forefront.

For manufacturers ready to explore how AI can improve their operations while protecting their workforce and data, James Moore’s team of experts stands ready to help facilitate that conversation and develop an implementation plan tailored to specific business needs.

Watch the full Moore on Manufacturing episode to learn more about implementing AI in your manufacturing operations. For personalized guidance, contact the experts featured in this discussion: Julie Kniseley, Tomas Sjostrom and Daniel Shorstein at James Moore & Company.