← Back to all Thoughts RSS Feed
LinkedIn icon September 01, 2026 • 2 min read • Originally published on Linkedin

The Real Bottleneck in AI Agents Isn't Reasoning, it's Context

The real bottleneck in AI agents today isn't raw reasoning capability; it's access to structured business context.

Google recently introduced the Open Knowledge Format (OKF), an open, lightweight Markdown and YAML-based standard designed to represent organizational knowledge so any AI agent, framework, or runtime tool can ingest it seamlessly. Today, engineering teams waste significant effort hand-crafting bespoke context pipelines for every internal tool. OKF replaces that friction with a portable, git-native, tool-agnostic knowledge bundle.

How OKF Compares to Existing Standards
To see where OKF fits into the AI architecture stack, it helps to contrast it with common alternatives:

OKF vs. Model Context Protocol (MCP): MCP provides the dynamic, client-server protocol layer—the live plumbing that connects an agent to tools, databases, and APIs. OKF operates at the representation layer: it standardizes the static, human-auditable knowledge assets (runbooks, enterprise definitions, domain constraints, schemas) that live inside version control before and during tool execution.

OKF vs. Raw JSON Schema Dumps: Raw JSON schemas describe low-level data structures, but they completely strip away semantic narrative, operational intent, and edge-case exceptions. OKF bridges that gap by blending human-readable Markdown with structured YAML metadata, providing both machine-parsable boundaries and the contextual nuance an LLM needs to interpret operational intent correctly.

Why Structured Schemas Matter: The Knowledge Graph Connection
This development aligns directly with a core architectural reality: neural networks require structured semantic backbones to remain reliable.

When agents interact with enterprise data lakes, APIs, or complex databases, letting an LLM infer or guess schema definitions on the fly leads to catastrophic runtime errors: hallucinated columns, malformed SQL queries, and broken API calls. Much like pairing ontologies and knowledge graphs with vector retrieval (GraphRAG), grounding agents in explicit, structured schemas anchors them in verifiable logic.

By turning institutional knowledge into a git-versioned, structured contract, standards like OKF ensure agents don't guess what your systems look like—they execute against validated boundaries.

Towards AI: Yor AI Agents Keep Forgetting Everything. Google Just Changed How They Remember

Github: Google Open Knowledge Format

Subscribe to the Technical & Geospatial Intelligence Briefing

Enjoyed this issue? Get the next one in your inbox. Every week, I synthesize the top cloud-native geospatial and data engineering articles, standout Google Research papers, and my latest essays.

Published weekly • Unsubscribe in one click • Zero spam.

Related Thoughts

Perspectives sharing related architectures, models, and domain context.

All Thoughts →
Sep 10, 2026 1 min read

AI Failure - It's All About Context

More often than not, AI agents end up failing because they’re drowning in their own context windows. One of the most...

Nov 06, 2026 6 min read

Why the Idea of a "Knowledge Graph Lakehouse" Has Me Thinking About Environmental Data

We've been working toward overhauling and reinventing our enterprise data platform, and this has me researching lots of...

Oct 20, 2026 3 min read

The Case for Neurosymbolic AI in Geospatial Workloads

My friend Bill Dollins recently published an excellent piece on neuro-symbolic AI and the geospatial workload over at...