Your AI agents are operating on 15% of the information they need

2026-06-15 · Source: Information and Enterprise Technology News | CIO Dive - Www.ciodive.com · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems, Cloud Computing & IT Infrastructure · Depth: Intermediate, medium

Summary

Max Romanenko of EDB, in an article published June 15, 2026, highlights that enterprise AI agents typically operate with only 15% of their context window dedicated to actual domain knowledge. This inefficiency stems from architectural decisions, where 25% is consumed by rules, 30% by orchestration, and another 30% by RAG chunks, leaving 85% of working capacity consumed before business knowledge is accessed. The article criticizes Retrieval-Augmented Generation (RAG) for its probabilistic nature, which is inadequate for deterministic enterprise decisions and contributes to agent unreliability. It proposes an architectural shift towards "intelligence at the data layer," where domain knowledge is structured, compressed (e.g., a 63,000-word book into a 20-kilobyte file), and served directly from the operational data layer. This approach, exemplified by EDB Postgres AI, ensures data sovereignty, auditable reasoning, and manageable risk.

Key takeaway

For CIOs overseeing enterprise AI deployments, prioritize architectural decisions that ensure context efficiency and data sovereignty. Your teams should measure the percentage of the context window reaching domain knowledge and ensure agents access live operational data layers, not just copies. Optimize for inference cost per useful decision, not merely token consumption, to build auditable, reliable AI systems that deliver on their promised productivity transformation.

Key insights

Enterprise AI agents are hampered by context window inefficiencies and probabilistic RAG, requiring an architectural shift to data-layer intelligence for reliable reasoning.

Principles

• Context window allocation is critical for enterprise AI architecture.
• RAG's probabilistic nature is insufficient for deterministic enterprise decisions.
• Data sovereignty is an architectural requirement for auditable AI.

Method

Proposes moving domain knowledge to the operational data layer, treating it as a structured, compressed artifact for direct machine reasoning, rather than relying on probabilistic retrieval.

In practice

• Measure context window utilization for domain knowledge.
• Ensure agents access live operational data, not copies.
• Optimize for inference cost per useful decision, not just tokens.

Topics

• AI Agents
• Context Window Management
• Retrieval-Augmented Generation
• Data Sovereignty
• Enterprise AI Architecture
• EDB Postgres AI

Best for: VP of Engineering/Data, AI Product Manager, Director of AI/ML, AI Architect, CTO

Related on AIssential

• Data Products: The Essential Context for Enterprise AI — Enterprise AI agent success hinges on robust, architected context layers, not just advanced models.
• The Next AI Bottleneck Isn’t the Model: It’s the Inference System — Misattributing AI system failures to the model rather than inference architecture is a common, costly mistake.
• Enterprise AI agents keep failing because they forget what they learned — Enterprise AI agents need structured, time-aware decision context beyond RAG's retrieval to prevent regression and ensure reliable, explainable actions.
• Mergeable by default: Building the context engine to save time and tokens — Peter Werry, Unblocked — Context engines optimize AI agent performance by providing highly relevant, curated organizational context, moving beyond simple data access.
• The Three Pillars of Agentic AI Systems — A Developer’s Field Notes from Build with AI 2026 — Effective agentic AI systems require structured memory, controlled actions, and self-correction for real-world operation.
• 🗞️ 1 million context window is now generally available for Claude Opus 4.6 and Claude Sonnet 4.6. — Advancements in AI focus on expanding context windows, optimizing inference, enhancing security, and automating research and training.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Information and Enterprise Technology News | CIO Dive - Www.ciodive.com.