Roadmap Towards Predictive Maintenance

Key insights from CEO Mohan Chaubal in an interview with Nilisha Kothari

Mohan Chaubal, CEO

In a recent episode of Reflexions, our CEO Mohan Chaubal discussed at length the opportunities and challenges in the integration of AI in Industrial IoT applications. Predictive Maintenance was a key topic of discussion. Here is an excerpt from his conversation with Nilisha Kothari.

Nilisha:: How do you look at the speed of advancements in AI technology and its adoption in the real world?
Mohan:: I think AI can be used in very effective ways, but the real world is not able to keep pace with the developments in technology. A few days ago, I was looking up Gemini for something related to IoT and I stumbled upon a statement saying Predictive Analytics for equipment maintenance is now an old concept! The truth is that AI research is advancing rapidly beyond predictive maintenance and exploring Agentic AI concepts like prescriptive maintenance. However, industrial adoption is still maturing. In reality, half the world still relies on reactive maintenance!

Nilisha:: What are the challenges in real-world implementation of predictive maintenance?
Mohan:: Only a small fraction of industrial businesses have implemented predictive maintenance at scale, some more companies are running pilot projects, but we don’t see large scale adoption yet. There is a reason behind this. Building AI model for predictive maintenance requires substantial historical data that many companies just don't have. Without sufficient,structured historical data, scalable predictive maintenance becomes extremely difficult.

Nilisha:: So how does one overcome this situation?
Mohan:: In order to implement AI profitably, the decision makers need to formulate longer term AI strategy and create plans that are divided into successive phases of implementation. The first step is to identify the operational and machine level parameters to be captured and to ensure that suitable mechanisms are in place for this purpose. The next step is to create data lakes of identified parameters along with the history of fault occurrences. Model training and actual predictions can begin once sufficient data is available. Refinements to the models may have to continue for a while for improving accuracy of predictions.

Nilisha:: This seems to be a long process and the outcome may not be predictable. How does a business justify the investment?
Mohan:: Yes, the process is long, but the benefits certainly justify the cost and effort. Data readiness is a strategic investment, not just a technical step. A lot more value can be extracted, beyond predictive maintenance, once historical data is available. For example, one may be able to understand the impact of seasonal or ambient conditions on equipment performance and identify ways to improve performance. Even as an organization is collecting data, they can expect some immediate benefits such as improving the outcomes of quality checks, understanding performance bottlenecks, and improving energy efficiency.

Nilisha:: You referred to Agentic AI a while ago. Can you tell us a bit more about that?
Mohan:: Yes, I was referring to the technology advancements beyond predictive maintenance. Once an organization has achieved predictive maintenance capabilities, they can go to the next step by implementing Agentic AI. After a maintenance need is predicted, Agentic AI can check the production plans and availability of spare parts, and schedule maintenance activity. If some spares are not available, it can also contact the vendors to receive quotes with price, delivery schedule etc. Purchase orders may also be generated automatically by integrating with the company's enterprise systems. But this is not merely about AI. There are bigger challenges to overcome including systems integration as well as governance and culture.
Of course, Agentic AI becomes possible only after the basic predictive maintenance systems are in place. We are currently helping businesses with the crucial steps to capture required data and to build AI models so that they can unlock measurable business benefits in the near and long term.