Why the manufacturers pulling ahead started years ago
Long before generative AI dominated manufacturing boardroom discussions, some organizations were quietly investing in something far less fashionable: data maturity. They established data governance programs, appointed Chief Data Officers, improved data quality and made information more accessible across the business. According to Jay Limburn, Chief Product Officer at Ataccama, those investments are now separating manufacturers successfully scaling AI from those still trapped in experimentation.
The reason, he argues, is straightforward. Artificial intelligence has not created a manufacturing data problem. It has exposed one that has existed for years. Organizations with trusted, well-understood data are moving from proof of concept to production because AI is working with information that is accurate, accessible and reliable. Those attempting to build AI on fragmented, inconsistent or poorly understood data are discovering that even the most sophisticated models cannot compensate for weak foundations.
“The biggest difference in terms of the patterns we see is that the companies moving forward are the ones that have mature data practices,” Limburn says. “Those organizations have spent the last five- or six-years putting data programs in place, appointing Chief Data Officers, making data accessible for analytics and improving the quality of their data. By far, they are the companies that have crossed the bridge from prototyping, experimenting and doing cool things with AI into delivering real business value, because the data is already in order. The outcomes are more accurate, they’re more trusted and they’re able to demonstrate value much more quickly.”
For manufacturers, the challenge is particularly acute. Decades of investment have created complex estates of enterprise applications, operational technology and production equipment that were never designed to work as a single data ecosystem. AI may be accelerating digital transformation, but Limburn believes the organizations creating lasting value are distinguished less by the technology they deploy than by the data foundations they established long before today’s AI boom began.
Data maturity changes the conversation
For many manufacturers, the challenge is not a lack of data but decades of accumulated complexity. Enterprise resource planning systems, manufacturing execution systems, historians, production equipment and cloud platforms have often been implemented at different times, for different purposes and using different standards. As a result, valuable operational information exists across the business, but bringing it together into a trusted resource for AI remains a significant undertaking.
Limburn believes the organizations making the fastest progress recognized this long before generative AI became a boardroom priority. Rather than waiting for AI to expose weaknesses in their data landscape, they invested in modernizing it through governance, quality improvement and greater visibility of the information they already possessed.
“There’s definitely not one-size-fits-all,” he says. “Companies have been investing heavily in platforms such as Snowflake and Databricks, moving their most valuable data from disparate sources into centralized cloud data platforms where it can be used for analytics. The organizations that have already started that modernization journey have partially solved the fragmentation problem. They’re now asking how they operationalize that data for AI. If you’re trying to deploy AI while still working with disconnected, poorly catalogued and poorly understood data, there’s a huge amount of risk associated with that.”
For Ataccama, that is where the conversation is changing. Historically, the company’s role was to help organizations understand, govern and improve the quality of their data before it was moved into modern cloud environments. Today, those same disciplines have become fundamental to successful AI deployment. Limburn argues that manufacturers no longer need data platforms alone. They need confidence in the information those platforms contain.
That is why Ataccama has focused on enriching emerging AI context layers with what it describes as “trust signals.” Rather than simply telling an AI model where information resides, those signals indicate where data originated, how reliable it is, how complete it is and whether it should be used in decision-making. As manufacturers move from experimentation to enterprise deployment, Limburn believes that ability to distinguish trusted information from questionable data will become just as important as the AI models themselves.
Trust has become the missing layer
For years, data governance was often viewed as a compliance exercise, designed to control access, enforce standards and satisfy regulatory requirements. AI is changing that perception. Manufacturers are discovering that governance is no longer simply about protecting information. It has become the mechanism that enables AI to use data confidently and responsibly at scale.
Limburn has seen a noticeable shift in the conversations taking place with customers. Historically, governance initiatives were driven by Chief Data Officers and specialist data teams, often with limited engagement from the wider business. Today, that relationship is reversing as operational teams recognize that successful AI depends on understanding, trusting and accessing the right information.
“We’re working with a very large food manufacturer that has been investing in data governance for around ten years,” he explains. “Historically, the governance team was pushing initiatives out to the business because it was important to protect data and meet regulatory requirements. The business didn’t always see the value. Now the business wants AI, but it also realizes it doesn’t have the data foundation to support it. Instead of governance teams pushing the message, the business is coming to them asking for help. That’s a fundamental change.”
That shift is encouraging organizations to rethink governance itself. Rather than acting as the “office of no,” restricting access and slowing projects, data teams are increasingly becoming enablers that help the business move faster. Automation is also changing the process. Activities that once relied on lengthy governance committees, such as classifying, profiling and understanding data assets, can now be completed far more efficiently, allowing organizations to build trusted data foundations without creating unnecessary bureaucracy.
Limburn believes that cultural change is just as important as technological change. The organizations gaining the greatest value from AI are those where business teams and data specialists work together from the outset, treating governance not as an obstacle to innovation but as the foundation that makes innovation possible. In his view, the conversation is already moving beyond data governance toward a broader concept of data enablement, where the objective is not simply to control information but to ensure it can be used safely, confidently and at the speed modern manufacturing increasingly demands.
Scaling AI means trusting every decision
The challenge becomes even greater as manufacturers attempt to move beyond tightly controlled pilot projects. Within a limited proof of concept, AI is often working with a carefully selected dataset that has been cleaned, validated and understood. Expanding that same application across multiple plants, business units or supply chains introduces far greater complexity, exposing inconsistencies that may have remained hidden during the pilot phase.
Limburn believes this is why so many promising initiatives struggle to scale. It is not because the AI models become less capable, but because they are suddenly expected to make decisions using data that varies significantly in quality, completeness and context.
“We used to live in a deterministic world,” he says. “People would look at the information and immediately recognise if something didn’t look right. If a purchase order normally requested ten rolls of steel but suddenly requested ten thousand, an experienced planner would question it. AI doesn’t think that way. It’s probabilistic. It makes decisions based on the information it’s given. If the underlying data is wrong, the AI has no reason to know that, and it will confidently produce the wrong outcome.”
That distinction has important implications for manufacturers. As AI becomes increasingly embedded within planning, procurement, production and customer service, organizations are placing greater responsibility on systems that depend entirely on the quality of the data they receive. The more those processes are automated, the less opportunity there is for experienced employees to identify errors before decisions are executed.
For Limburn, that is why data maturity can no longer be viewed as a technical objective owned exclusively by IT or data teams. Manufacturers that have successfully scaled AI have recognized that trusted data is now a strategic business capability. Technology will continue to evolve rapidly, but competitive advantage will increasingly belong to organizations capable of providing AI with information that is accurate, contextualized and trusted from the outset.
The foundations were laid long before AI
For manufacturers still struggling to move beyond isolated AI pilots, Limburn’s message is both reassuring and challenging. The organizations pulling ahead are not necessarily those with access to the most advanced algorithms or the biggest technology budgets. They are the ones that recognized, years before today’s AI boom, that data would become a strategic asset and invested accordingly. Governance, quality and accessibility were never simply data management exercises. They were the foundations of future competitiveness.
That investment is now paying dividends. As AI becomes embedded in engineering, production, procurement and supply chain operations, manufacturers with mature data practices are finding it easier to scale new applications, build trust among users and generate measurable business value. Those that delayed those investments are discovering that AI has not created new problems. It has exposed weaknesses that have existed within their data landscape for years.
For Limburn, the lesson is clear. The next phase of industrial AI will not be defined by increasingly powerful models alone, but by the quality, context and trustworthiness of the information that underpins them. The manufacturers creating lasting competitive advantage will be those that continue to treat data maturity not as a technology initiative, but as a core business capability. In many respects, the organizations leading the AI era won an important part of the race long before AI became the industry’s biggest conversation.
