Why Trusted Product Data Is the Missing Link for Enterprise AI

Enterprise AI does not begin with a model. It begins with confidence in the product data that feeds the model and the decisions built around it.

That distinction matters because companies can connect systems, complete integrations, refresh warehouses, and load dashboards without noticing that the information has lost its meaning. The most dangerous data outage may be the one no monitoring tool reports: every system appears healthy, but decision-makers are working from information they no longer understand correctly.

Enterprises are not struggling to build AI. They are struggling to make it matter. A CIO at a large global enterprise said data science teams had built more than 200 AI models, yet fewer than 10 reached production. McKinsey’s 2025 global AI survey found that AI use is widespread, but only about one-third of organizations had begun scaling AI programs across the enterprise.

The hidden risk inside product configuration

Product configuration sits at the center of this challenge. It acts as the connective tissue between engineering intent and the requirements used by manufacturing, sales, pricing, and service teams. When that logic remains accurate, it helps the business understand what can be built, sold, priced, and supported.

Configuration models often hold decades of accumulated knowledge. Regulatory requirements, manufacturing constraints, market-specific exceptions, and pricing logic can all become layered into the same model. If that logic changes or becomes inconsistent, the digital thread breaks no matter how well the surrounding systems connect.

Inconsistencies can quietly damage downstream processes and weaken AI recommendations. Without trusted configuration logic, AI does not improve decisions; it accelerates mistakes. A configuration may allow a company to sell something it cannot actually deliver, while another failure mode can block a valid combination. Both are expensive.

Enterprise modernization projects often focus on applications, infrastructure, and custom code while treating configuration logic as secondary. SAP S/4HANA migrations show why that creates risk: a new configuration engine can interpret long-standing rules differently even when the syntax remains identical.

Small differences can change the result. The way defaults are applied, the order in which constraints are evaluated, and the method used to resolve conflicts between rules can all change which combinations a model accepts, without anyone rewriting a single line of the rules.

From AI experiments to measured value

Cognite’s experience offers a practical example of what happens when AI work connects to measurable operational outcomes. The company announced that customers using its Industrial AI and Data Platform had unlocked more than $1 billion in verified, realized value.

The milestone arrived one year after Cognite publicly announced its 10-year Moonshot at its customer conference. The company has committed to delivering $100 billion in quantifiable value to Cognite customers by 2035, linking its success to the results customers can measure in their own operations.

“Documenting $1 billion in verified value in just twelve months is a testament to the deep collaboration among our customers, partners, and the internal teams at Cognite that support them,” Girish Rishi, CEO of Cognite, said.

Cognite paired its Industrial AI and Data Platform with a repeatable process for measuring return on investment. Its teams work with customers to identify the operational key performance indicators that matter most, then a dedicated value team tracks how those improvements affect the business.

During the 12 months after the Moonshot launched, Cognite and its customers identified billions in potential value and verified $1 billion in return linked to operational metrics used by industrial companies. Customers used AI and automated workflows to prevent costly asset shutdowns, restore operations after unexpected problems, and reduce operating costs.

Other customers gave field engineers context-rich operational insight, cutting time-to-decision from hours to seconds. That helped frontline workers make faster, better decisions, which supported higher production yields and lower costs. Increasing production through higher yields, faster cycle times, and fewer throughput bottlenecks accounted for 26% of the value customers realized, while 10% of value came from delivering solutions across complex operations faster than before.

Scale still needs people and systems thinking

Ignos, a strategic Cognite partner, connected 62 machines for one client in nine months and won back around 150 hours of machine uptime every day, Rolf Thu, its co-founder and chief technology officer, said. Ignos also launched “Mr. Off”—downtime with a face—to personify the disruption caused by unplanned downtime.

“When operators get the right data where they work, they create more,” Thu said. His point matches the broader lesson: AI creates value when trusted information reaches the people making high-stakes operational decisions.

COSMO Energy Holdings Chief Digital Officer Noriko Rzonca said industrial AI delivers true value when it helps companies solve those decisions and create tangible impact. She described Cognite’s $1 billion in realized customer value as proof that an outcome-first approach works at scale.

The platform’s usage also grew alongside the value claims. Customers exceeded 107 trillion contextualized time-series data points, up more than 39%, while data modeling instances doubled and AI token usage grew by more than 1000%. Since Cognite Flows Custom Apps launched earlier this year in May, customers have built more than 1200 production-grade apps to scale AI workflows.

Still, more data and more automation do not remove the need for judgment. Modern engineering projects combine human judgment with machine capability, but the final decision about whether and how to move forward remains a human responsibility.

Systems thinking becomes essential here. Engineers must decide how technical parts fit together, what work to give AI, what a person must check, when a specialist should decide, and when the problem itself needs to be redefined. The engineer remains a builder and specialist, but also becomes a coordinator of knowledge.

That role fits the idea of an “unbound” engineer: someone trained to understand complex systems, define and redefine problems, bring together the knowledge and tools needed to solve them, test whether solutions work, and take responsibility for the outcome. IEEE could support that shift through communities organized around problems that require several fields of expertise and practical standards for human-machine collaboration.

The path from AI experiments to dependable results runs through both trusted data and accountable people. Product logic gives AI a reliable foundation, measured outcomes show whether it helps, and human judgment determines how far the system should go.

Stay connected via Google News

Follow us for the latest AI News