June 24, 2026 Manufacturers Cannot Scale AI Without Trusted Data

What will manufacturers need to get data AI ready and unlock value?

By Michael Donahue

At a Glance

Every industry is accelerating investment in AI to get more efficient and almost everyone is still struggling to prove the ROI. In manufacturing, progress is even slower than the norm.

Even though 48% of manufacturers are piloting AI, only 10% have fully integrated AI into operations, that’s four points below the average of other industries combined, recent Grant Thornton research finds.

“Manufacturers are piloting. Competitors are scaling,”

– Grant Thornton 2026 AI Impact Survey

Many manufacturers are discovering that the biggest barrier isn’t AI models or ideas for innovation. Instead, it’s the quality, context, and trustworthiness of their own underlying data that limits the usefulness of AI. 

1. Data quality issues are widespread

Three of 10 manufacturers report experiencing data quality issues, recent Forrester research states. Poor data also limits technology adoption in manufacturing, the researchers concluded, and 91% of those surveyed said they faced barriers to being more data-driven. At the same time, manufacturing is increasingly awash with data. Factories generate massive volumes of machine and sensor data that can inform everything from predictive maintenance to supply chain optimization to whether factories should speed up production to meet existing demand.

But AI is only as good as the data that goes into it. It can only deliver value if the data is clean, complete, well-governed, and contextualized. On its own, AI doesn’t discern signal from noise. If it only receives outdated data, it can’t prioritize insights to inform manufacturing decisions today. It needs trustworthy information. Otherwise it can repeat flawed outputs.

2. Incomplete, inaccessible, and misunderstood describes much manufacturing data

Not all manufacturers are the same in terms of data and AI fluency. Manufacturers that are breaking into electric vehicles, for instance, have good quality data because they basically had to throw everything away and start from scratch.

But for the bulk of manufacturers, common data issues include:

All of these flaws mean that any AI will train on unreliable information and so will produce flawed insights. Without trustworthy, well-understood data, no AI initiative will deliver optimal value.

3. Getting data AI-fit requires intention

Many industries have long struggled with poorly kept and catalogued data. But AI has raised the stakes. If competitors make good use of the technology, they’ll quickly move ahead of those who don’t. Companies can start by asking themselves questions like:

What does AI-ready data look like for my operation? Being data-fit means you’ve cataloged what matters, labeled what’s useful, and can locate the most current version of any given asset—whether the data is structured in a spreadsheet or unstructured, like images of defects in manufactured products.

Do I have a lot of garbage data? Chances are, yes. This is true throughout all industries. We’ve all captured a lot of information and just kept it—long after its usefulness is over. This makes it harder for AI to find the truly important stuff. Get rid of the junk, reduce your cost, and optimize the performance of any AI engine.

Is my data trustworthy? Track the metadata, the data about the data, for lineage. Both agents and humans need to know where data comes from, how it was processed, what rules apply, and how it was changed over time. Tracking where data, or components, originate is especially critical for manufacturing that involves regulations and safety, such as food production. Without solid data lineage, all bets are off.

Do I understand my data? Use a catalog so everybody (machines and people) have the same definitions for data. Data also needs to be normalized. This way, one piece of data coming off one machine matches data meaning or showing the same thing that comes off another machine but may be named or stored differently.

What tools do I need to optimize my data or AI? Your data platform should connect across hybrid environments—cloud, on-premise, in spreadsheets—without needing to copy or move the data. An AI-ready platform must also operate across batch, real time and streaming data, sometimes all within the same use case. Manufacturing machines often like to hide their own data or charge to access it. You’ll want tools that allow you to go in and rip that data out and make it usable on the fly. With Pentaho’s automated data processing, the Association for Manufacturing Technology reduced query times by 80% for members to get access to data.

All in all, for AI to produce its best results, data needs to be “golden,” which means an authoritative datasets with full lineage, explainability, and embedded controls. Finally, you need a platform that makes sure that your manufacturing data stays private and secure and doesn’t end up in a large language model that then benefits a competitor.

Responsive, adaptive data platforms do the job

Every time there’s a major change in technology, such as the move to the cloud or off mainframes to PCs, other infrastructure needs change, too. For AI to fully reach its potential in manufacturing, along with practically every other industry, the foundation underpinning data needs to be adaptive and responsive.

Manufacturers are masters at squeezing out efficiencies and operating on margins that would sink other sectors. The value of manufacturing data—unleashed by AI—has more potential than ever to transform the industry in ways that stretch from Wall Street to main street.

About the Author:
Michael Donahue is a seasoned technology executive with over 25 years of experience bridging the gap between engineering, product management, and corporate strategy. As an “Enterprise Intrapreneur,” Michael specializes in scaling high-growth business units and architecting the data intelligence frameworks that serve as the foundation for modern enterprise execution.

His expertise focuses on the strategic evolution of data management, Sovereign AI, and ethical governance within highly regulated global sectors. Michael is a frequent speaker at innovation summits, focusing on the intersection of AI legislation and digital infrastructure. He holds an MBA from the University of Notre Dame’s Mendoza College of Business.

Subscribe to Industry Today