Skip to Content

AI can solve maintenance and quality challenges for manufacturers

Prasad Shyam

Companies know they need to integrate artificial intelligence into the manufacturing process, but the real challenge continues to be achieving it at scale. Scaling artificial-intelligence implementations beyond the proof-of-concept (PoC) level remains one of the biggest hurdles.

When the Capgemini Research Institute talked with manufacturers for the report Scaling AI in manufacturing operations: A practitioners’ perspective, the greatest amount of activity was in the maintenance (32%) and quality (26%) functions. The three most-implemented AI uses cases in operations were: intelligent maintenance, product quality inspection, and advanced simulation or digital twins.

It is not surprising that predictive maintenance is the number-one priority. In most cases, manufacturers can access significant amounts of historical and real-time data from machines to make reliable use cases, but it takes a change in mindset because many collected this data but did not look at it until it was too late.

The use of IoT sensors means manufacturers can access real-time data now. That opens up exciting possibilities. It means manufacturing execution can be made more efficient with fewer defects.

We worked with one automotive company on predictive models to identify robot failures. This proactive approach to maintenance improved quality and delivered real return-on-investment value.

Product quality is another area where manufacturers can improve efficiencies with AI. From stringent regulatory requirements to product specifications, non-compliance can lead to significant issues, from dissatisfied customers to fines and class-action lawsuits.

Detecting defects before they become a major issue is significant, and manufacturers should look beyond their own production lines when they compile data to find imperfections. For example, they should leverage the test data provided by suppliers. Manufacturers typically work with components from multiple sources, so assembling all the data from suppliers and the shop floor is critical for tracing back defects.

The challenge, however, is that suppliers keep data in different formats, so data engineering is required to combine it in a usable format. This is better than trying to make sense of a jumbled spreadsheet, and the ability to predict problems is worth the effort.

For example, one large manufacturer of computer equipment was able to predict a failure related to a heating issue before the product left the plant. It found that a cable connector was unable to withstand the required temperature, and that insight saved millions of dollars in warranty and recall costs. This type of root-cause analysis is vital.

Activating data in manufacturing is key to drive efficiencies and become a truly smart factory. To be successful, companies need to pick AI applications that can process real-time data from the shop floor and integrate with existing legacy IT systems. It will mean investing in data engineering, AI systems, and talent. But with the right governance and use cases, companies will find projects they can scale across multiple sites and factories.

Prasad Shyam is a Vice President, Insights & Data Global Practice at Capgemini. To learn more about how using data and analytics can improve business performance, contact him at