John S. Rinaldi, Real Time Automation (RTA) CEO & Expert in Industrial Networking & Open Smart Manufacturing. Author on Industrial Ethernet.
The Digital Transformation, and now AI, in manufacturing fails for one simple reason: inaccessible data, inconsistent and non-existent data modeling and a lack of governance. S&P Global/MIT (AI-specific) agrees, citing that over 40% of enterprise AI projects are abandoned due to poor data foundations and integration roadblocks. Predictive maintenance, quality analytics and digital twins depend on standardized, versioned and semantically consistent data models. Without this foundation, analytics become brittle, integrations multiply and comparisons of plant-to-plant operational efficiencies are problematic. According to McKinsey & Co., more than 70% of digital transformation initiatives fail to achieve their stated objectives or get stuck in pilot purgatory.
A data model in enterprise IT applications typically maps how data flows through a business’s management processes and entity relationships. Manufacturing data models, however, are far more complex and dynamic. Modeling physical assets—such as variable frequency drives, multi-stage pumps and robotic conveyors—requires capturing real-time telemetry, machine state, engineering units and operating context across thousands of highly variable machine configurations. Analytics teams spend 20% to 60% of their time manually cleaning, mapping and wrangling raw OT data before they can generate actionable insights.
Standardized Data Models
Smart Manufacturing systems that deploy company-wide, standardized data models are simpler to maintain, easier to extend and reduce the cost of integration. Standard data models allow AI systems to easily consume data, understand its context and identify potential actions to achieve more reliable operations. Plant applications that don’t rely on standards but instead rely on inconsistent or implicit data definitions are problematic, if not impossible, to maintain.
These aren’t just engineering conveniences; they’re national competitive necessities for the US manufacturing base. Yet industry adoption still lags behind awareness: while 77% of manufacturers recognize smart technology is vital to their market competitiveness, only half are willing to invest capital and barely 40% assign dedicated staff to execution. Recognizing this adoption barrier, the U.S. Department of Energy (DOE) funded CESMII (Clean Energy Smart Manufacturing Innovation Institute) in 2016 to target the root causes keeping smart operations out of reach: high cost, vendor lock-in, poor interoperability and the lack of a shared foundation for data-in-context.
CESMII delivers standardized, open information models (built on OPC UA foundations) that add context to data from assets and processes. Without context, data is often impossible to interpret. A value of 50, without context, could be interpreted as a temperature, a cycle time or a tank level. Context definitions must be reusable and shareable so data can move meaningfully between systems without custom and difficult-to-implement integrations.
CESMII initiatives require automation vendors to engineer and deliver more data-oriented products. Programmable Logic Controllers (PLCs) must evolve to add context where, today, none exist. Data aggregators, Supervisory Control and Data Acquisition (SCADA) and historians must store data with context and expose it in ways that anyone, not just domain experts, can access intelligently. Ideally, naming conventions must be standardized so the building blocks of analytics can be assembled and automatically recognized, allowing results to be presented in standardized ways.
Leveling The Playing Field For SMBs
Driving true interoperability is another central tenet of CESMII’s mission. Historically, plants only achieved cross-equipment communication by locking themselves into a single vendor’s ecosystem. Today, open standards from organizations such as CESMII and the OPC Foundation are breaking vendor lock-in, establishing a vendor-neutral framework for seamless data exchange across the plant floor.
Simplification is another driver of Smart Manufacturing. The trend is moving away from server-based solutions that require long-term IT investments and maintenance, toward more modular, hardware-oriented solutions that are simply installed and configured. HMIs are implementing technology that can use any screen. SCADA systems are becoming modular and containerized to ease deployment and maintenance and to increase reliability. New Historians are emerging that are modular and equipment-focused, an ideal starting point for small and medium businesses. As applications grow, their data can be aggregated into an enterprise-focused Historian. These solutions complement each other and let SMBs start small and grow as needed.
Where To Start: A Step-By-Step Guide
Traditional automation vendors are adapting their offerings by providing greater data context, open information models aligned with CESMII initiatives, Unified Namespace approaches and equipment-focused historians to reduce cost and complexity. Now is a great time for SMBs to invest in digital transformation, giving them a solid platform for continuous improvement. From my decades of experience, I find that factories that are the most successful in their digital initiatives follow a disciplined, four-step playbook:
1. Pick a single, high-value asset: Do not attempt to model the entire factory on day one. Select one critical bottleneck—such as a troublesome packaging cell, a high-draw compressor or a critical pump.
2. Standardize data at the edge: Extract raw tags from legacy PLCs and drives using simple edge gateways or modular protocol converters. Convert proprietary register maps into structured, contextualized data (OPC UA with standard companion specs or CESMII profiles) right where the data originates.
3. Implement a lightweight, asset-level historian: Log operational telemetry locally without commissioning heavy, server-based IT infrastructure. Then, prove that you can capture repeatable, contextualized trends for that single machine.
4. Expose the data via standard APIs: Connect the newly structured, edge data directly to basic visualization dashboards or analytics tools. Once you prove the value of standardized data on one machine, duplicate the model across the next line.
The Real Foundation Of AI Isn’t The Model
AI doesn’t create manufacturing insight; it consumes structured information and returns analysis. Feed it inconsistent, undocumented or implicitly defined data and the results remain unreliable no matter how sophisticated the analytics look. That’s why the foundation of Smart Manufacturing in 2026 isn’t the AI model; it’s the standardized data model that defines meaning, context, structure and access.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?