Agentic AI in Manufacturing: Foundations for Accelerated Adoption
A Reflective Start to 2025
I like to take time at the end of the year when customers are quiet and colleagues are away to spend some of my free time reflecting on accomplishments and working through the low-urgency but high-value tasks that I never seem to get to throughout the year. This ongoing work reflects both a professional responsibility and a personal commitment to staying informed. Similar to the end of 2023, I spent a good share of that time focused on investigating emerging technology in the generative AI landscape. This helped spur investments that I'm very excited for that are geared toward helping improve our team's workflows. During the break this year I dug deep, assessing what is real versus hype, and thinking through where we can apply the tangible advancements to improve the ways we work, optimize our methodologies, and ultimately deliver better results for our clients.
Fundamental Foundations in 2024
Over the past year, my team has refined our approach to digital transformation in manufacturing, focusing on the fundamentals that avoid risks and lead to successful results. These distill down to identifying business and process improvement opportunities, the implementation of effective solutions, and facilitating the changes that operations need to adapt to so they can realize value from their new solutions. Our team launched a comprehensive program through our DxOps Transformation framework where we clearly define what we do, the benefits our clients gain from our approach, and how the value will be realized from what we deliver. I’m also pleased to announce the launch of the RoviSys Insights blog—a space where our experts share actionable strategies and perspectives to support continued growth and improvement. As a system integrator, technology and platforms are a crucial factor to accomplishing manufacturers' goals, and our commitment to technical excellence has only expanded. Throughout 2024, our team has continued investing with our trusted vendors, gaining over 50 individual certifications across operations management solutions and MES vendor platforms.
Something Big is Coming
By now, there's a good chance you've heard the term "AI agents" or "agentic AI". Based on my own research and experimentation, the shock and awe that many experienced when LLMs and ChatGPT-type interactions were introduced was only the tip of the iceberg for what will soon be emerging. (And no, this article was not written by ChatGPT or the like... well, maybe I got a little assistance...)
Introduction to AI Agents and Agentic Frameworks
Agentic frameworks represent the next step in leveraging emerging technology to optimize and automate operational processes. Where LLMs have helped improve individual tasks and deliverables and have had a huge positive impact on people getting their work done, agentic concepts aim to coordinate and streamline the processes that even drive those tasks and deliverables. While they are gaining attention as a key innovation, their underlying principles may feel familiar to those experienced in leveraging large language models (LLMs) for decision-making and process optimization. For example, advanced prompt engineering with context chaining—where outputs are sequenced and informed by specific goals—mirrors the foundational logic of these frameworks.
What distinguishes agentic models is their ability to operate autonomously, processing complex datasets and identifying opportunities for improvement without constant human interaction. This evolution aligns with a broader trend in digital transformation, where tools are designed not just to execute tasks but to anticipate needs and deliver value proactively. For decision-makers, agentic frameworks represent a natural progression of the practices being implemented by industry innovators today.
Pesky Realities: Dependencies and Prerequisites
I can already hear it now; "Well great! Where can I buy one?" And trust me--someone will be happy to sell one to you. I hate to break the bad news--like we've learned in recent years from other technology innovations, agentic solutions are not magic beans or silver bullets that will solve all of your problems at the flip of a switch. In the context of manufacturing and industrial operations, for agentic models to succeed, a robust digital infrastructure is essential. Manufacturing Execution Systems (MES) play a pivotal role by organizing data, collecting context and data relationships, streamlining workflows, and preparing teams for digital integration. Without well-defined processes and an MES foundation, achieving seamless implementation of agentic models can be challenging.
Right Focus, Right Time
At the beginning of last year, my team's focus was squarely on, "How do we define solutions that will be useful in improving operational processes and decision making?" AI Agents and agentic frameworks were on our radars but were not the focus of our attention. Despite all of the interest around the *GPT hype, we kept our focus on defining how solutions improve operating processes. Fortunately--as it turns out--process and decision mapping is a fundamental enabler for applying agentic flows.
SIPOC Modeling and Process Structuring
The lack of well-defined processes in many manufacturing environments is a significant barrier to effective digitalization. Last year, our team adapted the SIPOC (Suppliers, Inputs, Process, Outputs, Customers) process modeling technique to help overcome this by breaking down workflows into manageable components and clarifying key decision factors. This structured approach provides a clear roadmap for implementing agentic models and maximizing their effectiveness.
Hype Versus Reality
If you are like me and have been swept up in the AI agent hype cycle, you've probably seen a few repeat examples of how these can be applied. And also like me, you're probably left wanting an example that is meaningful in a manufacturing setting--I mean there is not very much relevance between an agentic content marketing flow and a digital transformation program. Well, this is where the fun starts for me--and where the benefit of extended time out of the normal routine starts to pay back dividends. So, let's explore an example I've been thinking about.
A Common Challenge in Manufacturing
Daily/shift handoffs are critical yet often inefficient business process in manufacturing. Supervisors rely heavily on tribal knowledge and ad hoc communication, leading to gaps in information transfer. Processes are frequently undefined and managed on the fly, which can result in inconsistent decision-making and reduced operational efficiency.
Leveraging Agentic Models for Shift Handoffs
Agentic models offer a solution to these challenges by automating and streamlining several key aspects of shift handoffs:
- Data Aggregation: Data processing agent models can compile and summarize key performance metrics, ensuring that the incoming team has relevant and actionable information.
- Customizing Reports: Rather than rely on fixed-form reports, an agent model can adapt information based on daily performance factors, providing insights tailored to the specific needs of the shift.
- Automated Alerts: In cases where typical SCADA alarming is too rigid to provide meaningful alerts, agentic models can evaluate and highlight critical issues relevant to the current operating conditions, such as aggregated equipment downtime or resource bottlenecks, reducing the chances of oversight.
- Real-Time Detection: While constrained operations resources are sometimes barely able to repair what is broken, an agentic model can dynamically identify and measure emerging issues, allowing supervisors to make informed decisions without delays.
- Continuous Improvement: By integrating feedback loops, agent models enable ongoing optimization of processes and communication.
Practical Example: Enhancing Shift Handoffs
Imagine a manufacturing facility where agents organize daily shift handoffs. Before each meeting, the agents aggregate operational data, identify trends such as recurring downtimes, and prioritizes critical issues. By dynamically detecting performance bottlenecks, the agent ensures that supervisors are equipped with actionable insights, enabling seamless transitions between shifts. The things no one has time for now become possible. Want to review the control room call log? Digitize the information and feed into an agent to summarize the key issues and resolutions. This real-time capability not only enhances efficiency but also fosters a proactive approach to problem-solving. Will we soon be able to say goodbye to the days of digging through data or allocating a business analyst to prepare tailored reports?
Future Outlook
Looking ahead, I believe agentic models are positioned to become increasingly mainstream in manufacturing. For companies that continue investing in robust digital infrastructure and well-structured processes, these frameworks will enable more efficient operations, deeper insights, and proactive decision-making. While there’s no silver bullet, organizations that balance foundational excellence with emerging technologies can expect measurable improvements in responsiveness and productivity. Over time, agentic frameworks will likely play a pivotal role in evolving smart factories and more adaptive manufacturing ecosystems.
Conclusion
Agentic concepts hold significant promise for manufacturing, but their success depends on well-defined processes, a solid technology foundation, thoughtful implementation, and supportive adoption. By focusing on these fundamentals, organizations can unlock the tangible benefits of agentic models to drive operational excellence, adapt to market changes, and deliver greater value to their customers. In the end, it’s a deliberate blend of proven best practices and forward-thinking technology that propels true digital transformation.
