Demand-Driven Context: A Methodology for Coherent Knowledge Bases Through Agent Failure

2026-05-05 · Source: AI Engineer · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Intermediate, extended

Summary

Raj, a staff software engineer at IKEA, presented a workshop on "demand-driven context" for AI agents, addressing the challenge of institutional knowledge gaps in enterprise AI. He highlighted that while AI agents excel at reasoning and code generation, they struggle with domain-specific knowledge, leading to a delivery gap where 88% of companies use AI but only see 6% value creation. Current industry solutions like Retrieval Augmented Generation (RAGs) and Multi-Contextual Prompting (MCPs) often fail because enterprise knowledge bases are monolithic, containing outdated, unreliable, duplicated, and tribal knowledge (40%). Raj proposed a "pull strategy" where agents, like new employees, are given tasks and then iteratively pull necessary information, identify knowledge gaps, and curate new, structured context blocks. This approach, which has a preprint published in arXiv, aims to transform monolithic knowledge bases into usable context blocks, shifting agents from consumers to knowledge managers.

Key takeaway

For AI Engineers and ML Engineers deploying agents in enterprise settings, you should shift from pushing all available knowledge to a demand-driven, pull-based approach. This involves giving agents specific tasks, allowing them to identify and request missing institutional knowledge, and then having them curate and store that information. This method, especially when automated, can significantly improve agent performance and accelerate the transformation of your monolithic knowledge base into actionable context blocks, ultimately boosting the ROI of your AI initiatives.

Key insights

Enterprise AI agents struggle with institutional knowledge, necessitating a "pull strategy" for dynamic context curation.

Principles

• Knowledge curation should be demand-driven, not push-based.
• Agents can transition from knowledge consumers to knowledge managers.
• Automate knowledge gap analysis to identify critical missing information.

Method

Give agents problems they will fail, let them identify missing knowledge, provide the information, and have the agent curate and store the new context. Repeat this iterative process to build a robust knowledge base.

In practice

• Use GitHub repositories for managing curated knowledge blocks.
• Implement a meta-model to provide agents with a navigation map.
• Utilize a context gap scanner to automate knowledge base assessment.

Topics

• Demand-Driven Context
• Agent Knowledge Management
• Institutional Knowledge Gaps
• Context Gap Scanner
• Knowledge Base Curation

Best for: AI Engineer, Machine Learning Engineer, Software Engineer

Related on AIssential

• Knowledge Activation: AI Skills as the Institutional Knowledge Primitive for Agentic Software Development — Knowledge architecture, not model capability, is the bottleneck for effective agentic software development in enterprises.
• Dynamic Contextual Retrieval in Enterprise Analytics // Dirk Petzoldt — Effective enterprise AI analytics requires dynamic context retrieval and multi-agent systems, treating agents as consultants.
• AI Dev 26 x SF | Brandon Waselnuk: Building the Context Engine AI Agents Need — AI agents require a dynamic context engine to achieve organizational intent and avoid costly errors.
• 🔵 Linear says Issue Tracking is Dead — AI agents and context-driven workflows are fundamentally reshaping product development and traditional software roles.
• The context problem: Why enterprise AI needs more than foundation models — Enterprise AI requires deep institutional context to move beyond demos and deliver real production value.
• Context Engineering for AI Agents: A Deep Dive — Precise context shaping, not just quantity, is critical for robust multi-agent system performance and reliability.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AI Engineer.