AI pilot projects have stalled trying to scale, and unreliable AI results are unusable in mission-critical environments. We were all amazed by AI at first, with it being able to do incredible things to make our day-to-day lives easier. But the level of accuracy and trust needed for daily life vs industrial operations, where the risk of being wrong can be costly, is extremely different. Industry leaders are now running into the “Trust Gap”, where AI insights are being scrutinized and questioned. The industry is coming back to the data layer, realizing that it is an absolute need to stay competitive and survive in today’s market. While creating trustworthy data foundations, many are looking at a Unified Name Space (UNS) or knowledge graphs as the path forward. Those who can trust their AI will move faster than those who can’t.
How do I trust my AI insights?
Bridging the trust gap depends on a strong data foundation that is grounded in your industrial data, requiring:
- Data that is unified, curated, and governed in a central, accessible location
- Accessible real-time and historical data for AI to perform meaningful analysis and create insights
Certified datasets will be key to unlocking impactful value from agentic AI. These trusted, certified datasets ensure accurate data going into your agentic AI and create trust in the final results. This eliminates the data cleaning that accounts for ~80% of the time in most AI pilots, making it ready to use from the start.
How does a Unified Name Space (UNS) Fall Short for Industrial AI?
One way people are trying to tackle the Trust Gap is with the Unified Name Space (UNS). The Unified Name Space is a data architecture that organizes data hierarchically and provides the “now” of the plant. You can see real-time sensors and the status of various assets. An UNS provides live readings, such as pressure and temperature sensor data from a heat exchanger, and can work well to understand where a plant is in real-time. A UNS is built for a site-level implementation and works with OT data.
But what slows everyone down is the siloed nature of the data. UNS promises to centralize data and make it more accessible, but in reality, it fails to do so for 3 significant reasons. The hierarchical architecture is rigid and limits you to viewing the data through one lens. It relies on point-to-point data connections, and you’re restricted in how accessible the data is. On top of that, UNS is great for managing OT data but can’t handle ET and IT data, leaving a significant gap in the data your AI can access and reason over. And arguably, the biggest limitation is that UNS systems cannot access historical information to analyze trends.
What Is a Knowledge Graph, and why do I need It?
We’re watching the UNS crowd realize in real-time that a hierarchy isn’t the whole story. Why spend a year trying to solve a problem that is already solved? The future is here, and Cognite covers every UNS use case and a hundred others that aren’t covered by UNS.
- Bryan DeBois, Director of Industrial AI, RoviSys
Knowledge graphs are the next-step-up solution. Knowledge Graphs create a web of semantic relationships that contextualize data and accelerate data discovery, enabling operational troubleshooting in minutes and replicating success across multiple sites. UNS systems can coexist with knowledge graphs, but knowledge graphs simply have more capabilities and flexibility than UNS systems. Whereas UNS hierarchies are built for single-plant implementations, knowledge graphs are designed for enterprise-level scaling. Implementing a UNS across multiple sites means having to remap assets and sensors at every site, but knowledge graphs are flexible.
A knowledge graph takes it further, pulling in the surrounding context in addition to real-time data immediately. Using contextualized data from Cognite’s knowledge graph, a major refinery reduced RCA audit times from 60 days to 5 while identifying additional operational risks. To unlock this level of trustworthy insights from agentic AI, historical and contextual data from Knowledge Graphs are a must-have. UNS systems are limited to real-time plant data and cannot pull in historical data, which is crucial for driving decision-making. AI agents simply cannot learn from a system with no memory. Knowledge graphs quickly automate the discovery of the right data and context, grounding your decisions in your industrial data and eliminating hallucinations.
Grounding AI in industrial data so you can trust AI insights and decisions is exactly where Cognite thrives. We bridge the “Trust Gap” with a robust data foundation built with the industry-leading knowledge graph to keep you competitive against today’s market. It doesn’t have to be a struggle to bring transparency to AI insights and verify if its recommendations are the right decisions. The choice of data foundations isn’t about how much value a UNS brings to your site; it’s about how much further a knowledge graph can take your entire enterprise.