Too much siloed data, not enough action: How manufacturing leaders can turn operational data into competitive advantage with Industrial AI?

For CTOs, CIOs, and digitalization leads, the pressing need is to generate and collect data of good quality that is also accessible. Companies in discrete and process manufacturing industries, as well as those in the energy and utilities sectors, have been dealing with data for decades. However, their focus has been on operational activities at the expense of data maturity. 

“Usually, this wealth of data is used for very narrow purposes. Plenty of data may just sit in operational systems and logs, never been utilized for analytical purposes. But in the end, data should benefit people on the shop floor, in field work, in sales, in customer support, in product development, or the board room,” says Matti Partanen, Director of Service Solutions at Etteplan. 

Now, companies should aim to be data-driven and create holistic views by aggregating data from various operational data domains, including safety, quality, maintenance, aftermarket services, and the supply chain.

No wonder everyone is keen to try out artificial intelligence (AI), especially after the rise of generative AI. It promises to perform all sorts of data magic, leading to valuable outcomes. 

“Companies expect multiple benefits from leveraging AI, such as optimized performance, less downtime, reduced costs in materials and energy consumption, and better workforce productivity. Companies also try to find business-disrupting use cases, but often it is safer to start and build experience in internal operations,” Partanen tells. 

Industrial companies encourage and conduct AI pilots, in part due to their fear of missing out on the AI bandwagon. However, according to industry reports, a considerable percentage of AI initiatives stall and never move beyond the pilot stage. Why? 

Typically, AI is introduced without first establishing the foundations, which leads to disappointment. AI won’t function properly and deliver sensible results if there are bottlenecks in the availability, accessibility, and quality of data. 

According to Partanen, the key to success or failure lies in the company's data maturity. Therefore, companies should first assess how well digital technologies and automated workflows have been integrated into operations.

“I have witnessed industrial companies finding out the hard way about serious gaps in their data maturity. This is not surprising as production assets and operational systems are usually heavily siloed. Consequently, data is fragmented in various locations and formats. Data from relevant sources must be brought together, for example, to a data platform or edge that AI can access,” Partanen urges.

“There are organizational siloes as well. Sharing data that units possess has never been as relevant as it is today.”

Partanen recommends developing a proper data and AI strategy to start with. It must be closely aligned with the company's strategy to be successful. The strategy usually includes business opportunities and use cases. It also gives guidelines for the organization, leadership, data governance, technologies, privacy, and ethics.

“On the enterprise level, there is often a pitfall that business opportunities and use cases are stated at too high a level in the organization. However, they don’t fit in the factory’s real pain points or opportunities that well.”

It is easy to drown in building a perfect framework with all the bits and pieces of the data and AI strategy. To avoid this, companies should begin their AI projects with concrete use cases that are identified and operational pain points and build the strategic framework simultaneously. Pain points can be, for instance, yield loss, unplanned downtime, or excess energy usage.

“Having a clear use case is essential because you must steer AI tools to retrieve certain results and not just overwhelm end-users with too much information. For instance, you can task AI to investigate the causes of wear on a specific bearing in an industrial machine and combine this with human intelligence. Also, AI could aggregate information from several sources to a service operator in maintenance or aftermarket services and provide the findings in a format that is actionable and easy to comprehend,” Partanen says.

In many cases, the best approach is to use so-called AI agents, which are tailor-made and tasked with achieving a specific goal. Compared to traditional rule-based software and algorithms, they are much more efficient and adaptable.

Developing AI agents often yields the quickest wins, such as enhanced efficiency. They can also be integrated into internal workflows and processes to help human users make better decisions.

When AI starts to deliver results, people in the organization need to trust and understand AI-driven recommendations. That requires change management, transparency, training, and feedback loops.

“However, getting this far takes time and requires plenty of expertise. Success with Industrial AI requires a thorough understanding of the business, contextual knowledge of the industrial environment and engineering, IT expertise, and competence to work with data and AI models. Very few consultant companies working with the manufacturing, energy, or utility sectors can provide this. At Etteplan, we can, and have a proven track record to confirm this,” Partanen says.

Want to learn more about how to turn siloed operational data into real business value with Industrial AI? Download our free guidebook Create value with Industrial AI for practical insights, use cases, and a proven roadmap to scalable impact.