Is Your Shop Floor Ready for Intelligent Automation?

Speaker – Neerja - 01:20
Of course, Mandar. Manufacturers have been automating shop floor operations for years. 
Q: What makes intelligent automation different from conventional automation, and why is this shift becoming important now? Could you start us off here, please?

Speaker – Mandar - 01:37
Yes, sure. See, manufacturers have been automating shop floor operations for decades. PLCs, SCADA systems, robotics, and MES platforms have significantly improved productivity, quality, and operational efficiency. However, in the conventional automation part, they were designed to execute predefined tasks efficiently, based on certain fixed rules and programmable logic controllers. What makes intelligent automation different? You know, fundamentally, it has the ability to learn, adopt, predict, and continuously optimize the operations using your real-time data, AI, advanced analytics, and of course, the connected ecosystems within the plant. Intelligent automation leverages your latest technologies in terms of Industry 4.0 auto capabilities, industrial IoT, AI, digital domains, machine visions, advanced planning, and manufacturing control towers to make operations much more context aware and proactive. So,instead of simply collecting the data, the system understands what the data means and recommends or automatically executes the next best action. I can take an example here. See, for a machine breakdown scenario, AI can predict a failure day or weeks in advance, or let's say in case of a quality management solution, rather than manually monitoring the quality, vision systems can detect the defects in real time and prevent non-conforming products from moving further in the process. 
For production planning or production planning people or planners spending hours on sequencing decisions and coming up with best possible schedules, these days there are a lot of intelligent algorithms are available to dynamically optimize the schedules, based on demand changes, material availability, energy cost, and all other any or any other form of workforce constraints. Also, with Industry 4.0 technologies, the data has matured significantly, which is generated from machines, MES, ERP, supply chain systems, and connected assets. All this data has been brought in together through latest architectures such as unified namespace and manufacturing control towers, which creates a single operational view of your entire enterprise. Moreover, with Industry 5.0, the focus extends beyond automation for efficiency, the emphasis is more now on human-centric intelligent manufacturing. These days, a lot of AI co-pilots have been assisting operators, maintenance engineers, and supervisors and planners to make a faster decision and more informed decisions. See, for example, today I can quickly give you an example of let's say what is going to be my OE in shift one, shift two, and shift three. If there is going to be the deviation in OE in shift two, why there was a deviation in shift two? These kinds of questions I can quickly ask to AI copilots, and the answer would be available on my fingertips. So, in summary, my point is conventional automation helps manufacturers do things automatically, whereas intelligent automation helps manufacturers make better decisions automatically.

Speaker – Neerja - 04:50
Thanks, Mandar. And you know, intelligent automation, of course, has a lot to offer. 
Q: But where can agentic AI add value on the shop floor today. Could you walk us through a practical example from an alert to the action taken?

Speaker – Mandar - 05:07
Oh yes, absolutely. In fact, this can be an extension to our earlier topic. What we I just talked about. See, intelligent automation, like I said, at helping manufacturing in predicting the recommended actions, whereas agentic AI will take that step by orchestrating all those actions from the systems, people, processes, you know, and different technologies which is available in this manufacturing ecosystem. So, what happens today? Many plants, you know, you've seen they've generated 1000s of alarms and you know alarms and notifications and quality exceptions and managing alerts and whatnot, and there are a lot of professional deviations every day. So, the business value actually would be derived how you bring all of them together and contextualize them and then solving the problem. And that is where your agentic AI becomes a transformative technology, which will help you doing that. See, it can be you know imagined as a digital team member, which is not only analyzing the data, but it is putting that in context, understanding the reasons across multiple systems. It will also collaborate with humans as required. It will also check for you to know the business outcome, operational outcome from MES, ERP, SCADA, and historians, and any quality systems which are existing today within the plant ecosystem. I can give you a quick example. See, imagine a critical production line in a manufacturing automotive component industry where they are observing during the shift there are vibrations and some temperature sensors on a CNC machine, and there are some abnormal patterns which are getting generated. So, in a traditional setup, what would happen? A maintenance engineer will have to go look into the alert, investigate what has happened, check for some historical information, and then depending on the issue, it will search for a required spare part if it is to be replaced. Check in the inventory whether that spare part is existing today, and then he will have to schedule a downtime, and then you know he will have to coordinate with production and depending on these availability and priorities of different combinations, he will have to take a decision which may take hours and even days at times to do that. Now, with agentic AI, it is a digital team member, right? What it can do, it will continuously monitor the health. Immediately, it will correlate why this vibration has happened, what is the pattern it has followed. It will check the historical data. It will determine the probability of a specific component failure in this case, and when the machine would stop, let's say in 48 hours or so. So now it has got a complete traceability of information available and contextualize it better, and it can build that intelligence because of the orchestration capability it has. So, it will check for a production schedule from MES to APS, and then it will find out the least disruptive window, and then it will verify for the spare parts I be allocation in the ERP system. It will also review the technician schedule or any skill matrices which are required to do that kind of work, and it will impact the production and of course where there is going to be a revenue risk and potential downtime cost. So, the agent can provide a complete recommendation to that maintenance supervisor to take this action in a guided way. How to replace that, you know, bearing replacement recommendation, change the window of maintenance, spares which are to be brought in automatically. What is the estimated downtime? All of that can take into cognizance and then provide a recommendation and take an action also. Now, like I said, if it's human in loop, it will ask you if you want to take this action within so and so time, or you want to kind of, you know, want me to generate only at an alert level. So, all those decisions can be done with an agentic AI. So, the key takeaway is agentic AI is a technology that basically will close the gap, converting all your intelligent insights into a measurable business outcomes. That is where agentic AI would add value to the overall transformation journey.

Speaker – Neerja - 08:43
Yeah, and it can very much become a team member, like you said, Mandar. But before manufacturers give AI a role in you knows production decisions, what foundations must be in place across shop floor data, MES, and other operational systems, cyber security governance, and human oversight.

Speaker – Mandar - 09:04
Yeah, that's a very interesting question. Any form of AI has tremendous potential, right? No manufacturers should think about automating a decision making before establishing the right digital foundation. So, the first foundation I would talk about is the trusted and contextualized shop floor data, and this data today is available within the shop floor ecosystem in the form of PLCs, SCADA systems, historians, quality AES, and ERP applications. And most of this data is fragmented, inconsistent, and sometimes it is unreliable too. So, if these data or tags, equipment tags, which is not standardized or asset hierarchies are incomplete, or the process data lacks the contextualization. Naturally, AI will struggle to generate a meaningful recommendation, and that is reason you will see a lot of new technologies like you know unified namespace, industrial data platforms, and manufacturing control towers. They have been brought into the picture to create a single source of operational truth. Across the plants and business functions, the second important foundation in this case is MES. AI cannot effectively drive the outcomes if it is disconnected from core manufacturing process, which is the job of MES to ensure you know all these things are tied back together with respect to the production execution, quality, genealogy, traceability, and work instructions. So, in case if there's going to be a machine failure, one need to understand its impact on production schedules, customer orders, you know, inventory availability, and the real value emerges when AI operates within the context of end-to-end manufacturing processes rather than an isolated data streams, where AES is doing this job of binding all these data points together across different functions, which I talked about. The third important foundation is cyber security and operational resilience, right? So, we know these days all the manufacturing ecosystems are increasingly connected, and naturally this leads to you know exposure to a larger attack surface. You know the cyber-attack surface, AI systems consuming this information and analyzing this data sometimes also influencing lot of operational decisions. So, it needs to be taken care of properly through a proper secure IT-OT convergence, providing right access to the role-based access controls and data protection, network segmentation across ITO team, and continuous monitoring. The fourth important foundation is governance and accountability. See, most of the executives are concerned not because of whether AI can make a proper recommendation, but whatever recommendations have been made by AI, are they trusted? Can they be trusted? So, manufacturers need clear governance mechanisms and frameworks who can own the decisions. How you know those recommendations can be validated? Which are those actions which can be automated and where human approval remains mandatory? 
So, in essence, the manufacturer should view AI adoption as a maturity journey, and they can also utilize this journey and take it towards agentic AI-driven orchestration with appropriate human oversight, but the foundation has to be right from basis those four pointers what I just talked about, which will unlock not only the productivity gains but also the trust required to scale the AI across enterprise responsibly.

Speaker – Neerja - 12:21
Great, Mandar. I'm sure this information is extremely useful for everyone listening in. And I will ask you: 
Q: For manufacturers ready to begin with intelligent automation, where should they start, and how can they assess readiness, choose a first use case, measure the results, and ultimately expand what works across production lines and plants.

Speaker – Mandar - 12:45
Yeah, that's a very good question. Again, from my experience, I can definitely answer this. See, data readiness is the first and the foremost step when you are looking at any such kind of initiative. So, it will basically enable your operational data availability. How reliable is your data, and can it be contextualized? Is it accessible? All these points are required to be considered from the evaluation’s perspective. Then, next comes your process maturity. Has there been any standardization and measurable outcomes? What you want to derive out of that, from that perspective, is the process really standardized, or do I need to bring in some changes to do that. Third is the technology landscape. So, what kind of systems do you have today? Whether you have MES, ERP, quality, all of that, and of course, what is the governance mechanism from the people perspective? Do you have required skills and people who are taking ownership to drive this kind of a transformation? And like I said in my earlier answer, also about the cyber security and governance aspects, which can basically drive the AI transformation responsibly. Once it is established, then the next part, which is very important, is the identification of the use case. Which use case will yield a required business value. Now that use case can be tied back to your predictive maintenance. It can be your quality optimization through Vision AI solutions. I want to do production planning and scheduling first, or energy optimization, or OEE improvement. So, these are all critical areas for any manufacturing industry. If you go for, and then on the basis of this, which can yield me a quick business value and it will be a good business case for me to take up, I can identify that use case and with a tangible business outcome, then you measure the business value by implementing a pilot. After this, you pick up one specific use case. Say, for example, I want to improve my OEE, and what impact it can really generate because OEE is a larger KPI for any industry, which would address your availability, productivity, and quality altogether. So, I may pick and choose that, implement it on one line, and then I start extending this to other lines depending on the business value I have been able to generate with one line. But I have a roadmap for scaling it across multiple lines, and that is how we come to a complete industrialized view and taking it to the next level of you know implementation. So, all I'm trying to drive a point home here is, you know, rather than deploying a disconnected solutions plant by plant, or you know, just having a platform would not help. It needs to go in a structural way from the process perspective, from the technology perspective, and of course from the people perspective, wherein you can have the latest and greatest of the tools used, which is available in the in the industry, just an example: industry data platforms, unified namespace data architectures, and with this, you will be able to scale your implemented use case very quickly across multiple lines, plants, regions, and with significantly lower cost because a lot of standardization and best practices has been brought into these new architectures, which I just talked about. 
So, in a summary, I can say the journey should follow a very simple roadmap, checking with the assessing the readiness, and then prioritizing those high-value business use cases, which is basically linked to your business outcomes, and finally defining those scalable digital foundations, and then expand it across planets to generate maximum value, so this is going to be a very pragmatic approach, which I would like to share based on my experience. And there is always a transformation which leads from industry 4.0, 5.0, which I'm envisioning in various pursuits also, and industry is already moving in that direction. So that would be my final take on this.

Speaker- Neerja - 16:18
Mandar, there is so much to learn from what you said, thank you for being such a generous source of information today, and we really appreciate you sharing your knowledge with us.

Speaker – Mandar - 16:29
Thank you, thank you, Neerja.

Speaker – Neerja - 16:30
As we wrap up, I will leave you with one simple takeaway: start with a real operational problem, connect the right data and systems, and keep people in control of critical decisions. Thank you for listening to Tech Lyceum. Until next time, stay curious.

Running Music – 16:48
You were listening to Tech Lyceum, a podcast from Birlasoft.