KoGen - 3 layers

IBM technology partner, Equitus.us's -  Kogen (Knowledge Generation) on its KGNN (Knowledge Graph Neural Network) architecture can significantly improve how IBM Power 11 users interface with the three layers of AI: Generative, Agentic, and Iterative . The core of this improvement is the Knowledge Graph , which provides a structured, context-rich, and verifiable data foundation that overcomes the limitations of large language models (LLMs) running on the Power 11 system.

---

How the KGNN Architecture Enhances Each AI Layer

1. Generative AI

Generative AI, such as LLMs, excels at creating content but is prone to "hallucinations" and struggles with factual accuracy, especially when dealing with a company's private, siloed data. The Kogen KGNN architecture addresses this by acting as a powerful Retrieval-Augmented Generation (RAG) engine. It ingests and unifies disparate data sources from across the enterprise—documents, databases, emails, etc.—and automatically organizes them into a knowledge graph. When a user queries the Generative AI, the system first retrieves precise, factual information from this graph. The LLM then uses this verified context to generate a truthful and accurate response, ensuring the output is grounded in the user's own data. This is particularly valuable on IBM Power 11, which is designed for on-premises data sovereignty.

2. Agentic AI

Agentic AI, or AI agents, are designed to perform multi-step tasks autonomously. To do this, they need to reason and make decisions based on a deep understanding of relationships between entities, data, and processes. The Kogen KGNN provides the essential structured context for this. An AI agent can traverse the graph to identify dependencies, understand complex relationships, and plan a series of actions to achieve a goal. For example, an agent tasked with diagnosing a supply chain issue could use the knowledge graph to link a product's manufacturing date to a specific raw material supplier, a shipping route, and a customer complaint, then use these relationships to propose a solution. This prevents the agent from making uninformed decisions and enables it to perform complex, trustworthy actions.

3. Iterative AI

Iterative AI refers to a system's ability to continuously learn and improve. The Kogen KGNN architecture facilitates this by creating a feedback loop where the knowledge graph itself can be dynamically updated based on the outcomes of agentic actions. The system can log successful and failed actions and their consequences directly into the graph. This allows the AI to self-correct and refine its strategies over time without extensive human intervention. As new information, behaviors, or business processes are introduced, the knowledge graph evolves, ensuring the AI remains relevant and effective. This continuous learning process, powered by the structured data in the knowledge graph, makes the entire AI system more robust, adaptive, and intelligent over time.