Q&A: How Contextualized Data and AI Agents Enhance Manufacturing Automation

AI agents in industrial applications have garnered attention as the must-have emerging technology. Pundits tout AI agents for their genius and utility in creating new business opportunities, new products and services: AI agents stabilize production quality. AI agents improve yield and minimize the need for manual inspections. AI agents have 24/7 capability.

AI agents are a progression from generative AI, which turns large amounts of data into actionable knowledge. In contrast, AI agents are probabilistic technology with high adaptability to changing environments. They are autonomous in their decision-making processes because they use available tools such as external data sets, online searches and APIs to learn, reassess, update and adapt to user expectations over time.

Is an AI Agent Coming for Your Job?

Business leaders in a recent survey conducted by Forrester Consulting expressed concern about the lack of synergy between their design and manufacturing teams. Improvements in operational efficiency and the ability to advance the pace at which new products get to market are impeded by not making digitalization and organizational changes, revealed respondents in the survey.

For manufacturers that are unprepared, making use of industrial data at scale can be overwhelming. Generative AI models using text, images and code are already transforming organizations by turning large volumes of data into actionable knowledge, helping employees work more. By extension, AI agents could further enhance performance by improving the synergy between their design and manufacturing teams.

Yet, until industry stakeholders are confident that their use cases and value have demonstrable gains (in efficiency and cost savings) and that they’re covered against things that can go wrong (litigious issues and copyright infringements), they will not be fully ready to assign tasks to advanced technologies such as agentic AI.

Although autonomous, AI agents are arguably in the early days of development, pointed out Aron Semle, chief technology officer and co-founder at HighByte, an industrial software company in Portland, Maine. They require feedback mechanisms, notably human feedback, to perform according to human expectations and standards. Humans likely won’t be taken out of the loop anytime soon.

Prepare for the Leap to Data Adaptation in the Engineering Flow

In the Q&A that follows below, Semle shares more insights into contextualized data and how the use of agentic AI can transform manufacturing and mechanical engineering tasks.

Machine Design: What are the benefits of contextualized data production?

Aron Semle: Historically, manufacturing data has stayed within the factory. When you're on-site with a CNC machine, you can see both the data and the machine in action, making it easy to interpret processes—including how the data is measured, when it’s valid and what it means in real time. But when you're in a data center, far removed from the factory floor, it’s more difficult. Context is what makes data meaningful. It allows people and AI applications to make sense of the information, ensuring it is actionable rather than just raw numbers. Without context, data is just noise. With context, it becomes valuable.

MD: Despite the potential, AI agents pose certain risks. What are the technical limitations?

AS: The biggest risk of data without context is misinterpretation. When manufacturing data is separated, it becomes much harder to validate and understand. This can lead to incorrect conclusions, inefficiencies and even costly mistakes in decision-making. For AI-driven analysis, the risk is even greater. First, if you don’t get your data ready for AI, you will be at a big disadvantage both in optimizing as well as in recruiting new talent.

And secondly, without data context, AI models may generate inaccurate predictions, misidentify anomalies or fail to optimize processes effectively. In manufacturing, where precision is critical, these errors can result in unnecessary downtime, defects or lead to negative output performance.

MD: What are AI agents and what does it mean specifically for the design and mechanical engineering community and for manufacturers in general?

AS: AI agents are a bit of a buzzword, so if you ask different people, you’ll get different answers. Unlike using an LLM, where you can guide and refine responses in real ti,me, agents don’t have that kind of direct interaction. You give them a task, provide access to the right data (with context) and they run on their own.

For manufacturing and mechanical engineering, this could mean something like, “Monitor this process and flag anything that looks off.” Given the critical nature of manufacturing, the first wave of AI agents will likely be operator assistants—observing and notifying but not making direct changes to processes or set points. Another big use case is operational planning assistance, like “Reschedule production based on current line conditions to ensure orders X, Y, and Z go out today.”

The biggest challenge right now is that AI agents don’t always produce consistent results, meaning if something starts going wrong, the agent cannot course-correct and may drift further into errors or hallucinations. That’s why AI agents won’t take humans out of the loop.

MD: Are there any specific takeaways for the engineering community? What will stay the same, what will change in their systems?

AS: Generative AI and LLMs are already transforming engineering tasks like coding and logic development, helping people work faster and more efficiently. These tools are an opportunity—not a threat—as they won’t replace engineers but will enable them to do more.

The biggest shift will likely come from AI-driven operator assistance in manufacturing, which could reignite interest in the field. To stay ahead, the community should focus on getting data ready, as AI is only as good as the data it’s learning from. It’s critical that you look at the data in your factory today and start thinking about how you can make it accessible to people—at the edge and in the cloud. While AI will automate many tasks, the need for human expertise isn’t going away. Staying involved in the physical side of manufacturing, where AI has limits, will make engineers even more valuable.