By Ryo Kaneko, Director of Innovations, NEC X 

AI agents have made rapid progress. They can write code, automate workflows and handle increasingly complex tasks. On the surface, they appear close to real-world readiness. 

But in practice, reliability remains a challenge. 

Performance degrades as tasks become longer, outputs vary even with identical inputs, and human oversight is still required. For enterprise environments, this lack of consistency is a critical barrier.  

This is not a temporary limitation. It is a structural one. 

The Core Problem: AI That Reasons From Scratch

Most AI agents today are designed to interpret context and infer decisions in real time every time. 

This approach enables flexibility, but it also introduces instability. Because reasoning is probabilistic, outputs can fluctuate. As workflows become more complex, small errors accumulate, reducing overall reliability. 

This is why many AI systems perform well in controlled scenarios but struggle in production environments. 

To move forward, we need to rethink a core assumption: that AI must “think” from scratch for every task. 

A More Practical View: Real-World Problems Are Structural

In real-world operations, most problems are not entirely new. 

They may appear different at the surface level, but their underlying structure, the relationships between conditions, signals, and outcomes, remains consistent. 

What changes is language. What stays constant is structure. 

This distinction is critical. Many enterprise workflows do not require continuous reasoning. They require the ability to recognize patterns that have already been seen before. 

However, most AI systems today operate primarily on unstructured text, which is inherently ambiguous and inefficient to reuse. 

The Architectural Shift: From Language to Structure-Based AI

To address this, AI systems must move beyond language and operate on structured representations of knowledge. 

By converting inputs into graph-based relationships, problems can be expressed as networks rather than text. This allows systems to focus on the essential structure of a situation instead of reinterpreting it each time. 

Graph Neural Networks (GNNs) play a key role here. Rather than analyzing meaning in isolation, they learn patterns across relationships, identifying which structural configurations correspond to specific problems and solutions. 

This enables a new architecture: 

• Inputs are transformed into structured representations  

• Knowledge is stored in a shared graph-based memory  

• Patterns are matched using graph-based models  

• Execution agents act based on these structured insights  

In this model, reasoning is no longer recreated repeatedly. It is grounded in accumulated knowledge. 

Memory itself becomes an active component of intelligence: interpreting, contextualizing, and guiding decisions. This creates a stabilizing layer that supports and improves AI agents over time. 

From Intelligent Systems to Reliable Infrastructure

As AI adoption accelerates, the focus is shifting. 

The question is no longer just how intelligent a system is, but how reliable it is in real-world use. 

Structure-based AI addresses this directly. By reducing reliance on repeated probabilistic reasoning, it improves consistency, accuracy, and scalability. It also allows systems to continuously improve as new data is incorporated into their knowledge base. 

This marks a broader transition: 

• From AI that reasons every time to AI that recognizes and applies structure 

• From standalone tools to infrastructure for operating and scaling knowledge 

For AI to succeed in enterprise and mission-critical environments, this shift is essential. 

At NEC X, this perspective reflects a broader focus: building AI systems that are not only powerful, but dependable and designed for real-world deployment and measurable impact. If you’re interested in building such a solution, learn more about our Elev X! venture studio programs here.