The Dirty Secret Stalling Industrial AI: Why Data Is the Real Bottleneck

PEWAUKEE, WI – June 17, 2026 – The narrative surrounding Industrial AI is one of transformative potential—a world of self-optimizing production lines, predictive maintenance that eliminates downtime, and unprecedented efficiency. Yet, for many manufacturers, this vision remains stubbornly out of reach, stuck in an endless cycle of pilot projects that fail to scale. The culprit, however, isn't a flaw in the algorithms or a lack of ambition. It's a far more fundamental and pervasive problem: the chaotic state of operational data.

While AI is often heralded as the 'brain' of the modern factory, it's being fed a diet of digital junk food. According to a recent survey, a staggering 73% of manufacturers report that poor data quality prevents them from gaining actionable insights. Industry analysis from firms like IBM corroborates this, suggesting that as much as 90% of all data collected on the factory floor goes completely unused. This is the critical, often-overlooked challenge that John Rinaldi, a recognized expert in Operational Technology (OT) and CEO of Real Time Automation (RTA), plans to tackle head-on at the upcoming Automate 2026 conference in Chicago.

In a session titled "Beyond the Algorithm: Standardized Manufacturing Data Models as the Foundation for Scalable Industrial AI," Rinaldi will argue that before companies can achieve AI-driven transformation, they must first address the data foundation. His message is a pragmatic and urgent call to action: stop chasing shiny AI objects and start cleaning up the digital shop floor.

The AI Scaling Wall: A Foundation of Sand

The reason so many Industrial AI initiatives hit a 'scaling wall' is that they are built on a foundation of sand. Decades of organic growth, mergers, and proprietary systems have left most factory floors with a tangled mess of data. Programmable Logic Controllers (PLCs) from different eras and vendors use inconsistent tag names—what one machine calls 'Motor_Speed,' another might label 'RPM_1A.' Units are often implicit, and crucial context about processes and equipment states is left undocumented, locked away in the minds of veteran engineers.

This creates what Rinaldi calls the "Jenga problem" in legacy data historians—the traditional databases used to store factory floor data. These systems are often rigid and brittle; attempting to change, integrate, or scale them is like pulling a block from the bottom of a Jenga tower, risking the collapse of the entire structure. The result is data that is siloed, ambiguous, and untrustworthy.

"Engineers no longer struggle to collect data; they struggle to make it useful," Rinaldi stated in the announcement for his talk. This sentiment is echoed across the industry. Data scientists working in manufacturing often confide that they spend up to 80% of their time on data cleaning and preparation, a process of digital archaeology to make sense of raw, context-free information. This inefficiency is a primary reason that Gartner reports over half of all AI projects never make it to production, with data issues being a leading cause of failure. It's a systemic drag on innovation that costs organizations millions in wasted effort and lost opportunity.

Beyond the Algorithm: The Blueprint for AI-Ready Data

According to Rinaldi, the solution lies not in more complex algorithms, but in a disciplined, strategic approach to data management through standardized data models. This isn't just about creating tidy spreadsheets; it's about building a universal language for manufacturing operations that both machines and people can understand. This blueprint for AI-readiness consists of several key components:

• Rich Metadata: This is the 'data about the data' that provides essential context. Instead of a raw value like '70.5,' a standardized model ensures the system knows this value is 'Temperature' from 'Sensor_ID_25B' on 'Extruder_Line_4,' measured in 'Celsius,' with the last calibration date noted.

• Consistent Schemas and Naming Conventions: By enforcing rules for how data is structured and named across all assets and plants, manufacturers can eliminate ambiguity. This ensures that data from different sources can be accurately aggregated and compared, forming a reliable 'single source of truth.'

• Standard APIs: Application Programming Interfaces like OpenAPI (for synchronous data requests) and AsyncAPI (for asynchronous event streams) create a stable, documented contract for how data is exchanged. This allows AI platforms, enterprise systems, and new applications to plug into the operational data stream without custom, brittle integrations.

This approach aligns with broader industry movements toward interoperability, such as the adoption of the OPC UA standard, which provides a framework for secure, contextualized data exchange. By implementing these principles, manufacturers can transform their chaotic data streams into structured, contextualized, AI-ready information.

A Pragmatic Path Forward with Modern Tools

As a 35-year veteran of industrial connectivity and the author of six books on topics from OPC UA to Modbus, Rinaldi brings a uniquely pragmatic perspective. His company, Real Time Automation, is championing this data-first approach with its 'modern historian' solutions, which will be demonstrated at Automate 2026.

These modern systems are designed to directly address the shortcomings of legacy historians. They are built for flexibility and scalability, solving the "Jenga problem" by creating a more modular and robust data architecture. Furthermore, Rinaldi highlights the need to escape the "subscription pricing trap" of many enterprise software solutions, which can create prohibitive recurring costs that stifle broad adoption. RTA’s approach suggests a focus on providing value and a clear ROI for plant engineers and system integrators, rather than locking them into ever-increasing fees.

By offering gateways and software that bridge the gap between legacy machinery and modern enterprise systems, the company provides a practical pathway for manufacturers to begin standardizing their data without a complete rip-and-replace of their existing infrastructure. This allows them to build the necessary cybersecure and contextualized foundation for AI incrementally and strategically.

From Data Janitor to Strategic Asset: The Business Payoff

The strategic implications of getting this data foundation right are immense. When data is structured, contextualized, and reliable, it transitions from being a maintenance headache to a core strategic asset. McKinsey estimates that effective data management can unlock productivity gains of 20-30% in the industrial sector by enabling advanced analytics and AI at scale.

With a solid data foundation, the promises of Industrial AI become achievable. Predictive maintenance models can ingest clean, consistent data to accurately forecast failures, dramatically reducing unplanned downtime. AI-driven process optimization can analyze reliable real-time data to fine-tune production parameters, boosting yield and cutting energy consumption. Quality control systems can leverage standardized data to identify defects with greater precision, reducing waste and protecting brand reputation.

Ultimately, the journey to a smart, resilient, and competitive manufacturing future doesn't start with a complex algorithm. It starts with the disciplined work of building a data infrastructure that is ready for the challenge. As leaders like Rinaldi emphasize, mastering the data is the first and most critical step in mastering the future of manufacturing.