Your AI tools are only as smart as the context you give them. That’s exactly the gap Google’s Open Knowledge Format — OKF — is designed to close.
Announced by Google Cloud in 2026, OKF is an open specification for organizing business knowledge into a format that both AI agents and humans can read and use directly. “Open specification” means it’s a publicly available standard — no license fee, no proprietary platform, no SDK required to adopt it.
The core idea: take your company’s knowledge — metric definitions, support playbooks, data schemas, operational procedures — and package it in a structured folder of plain text files. AI tools read those files directly. So does your team. Same file. No translation layer.
The Problem It’s Solving
Most businesses have a knowledge problem. Your pricing rules live in a Google Doc. Your refund policy is buried in an email thread. Your definition of “active customer” differs between your finance team and your sales team.
When you connect AI to that mess, the AI reflects the mess. Inconsistent answers. Outdated facts. Confident AI responses that are just wrong — because the model grabbed stale information from two contradictory sources.
OKF gives your knowledge a structured, permanent, version-controlled home.
How It Works
An OKF bundle is a folder of Markdown files. Markdown is a plain-text writing format — you can open it in any text editor, it renders on GitHub, and any AI model can read it directly. Each file represents one concept: a metric definition, a product specification, a runbook for handling a customer complaint.
At the top of each file sits a short YAML block. YAML is just a few lines of structured labels — like tags and timestamps — that tell any system what type of thing this is. Below that: plain prose.
No database. No proprietary account. The bundle lives in a folder, travels as a zip file, stores in Git alongside your code, and works with any AI tool on the market.
Google published the full spec on GitHub — free, open source — along with sample bundles covering GA4 e-commerce data, Stack Overflow’s public dataset, and Bitcoin transaction data. You can see exactly what it looks like before writing a single file.
What Businesses Are Doing With It
Defining metrics once. A finance team packages their approved definition of “weekly active users” — the formula, the data source, the exclusions, the owner — in a single OKF file. Every AI agent and every analyst pulls from that same file. No more conflicting numbers in executive reports.
Automating customer support. A customer service team writes their refund rules, escalation procedures, and subscription tiers as OKF concepts. The support AI reads those files instead of guessing. Fewer wrong answers. Fewer escalations. Less time training and re-training the bot every time policies change.
Making data usable for AI. Engineers map table schemas, join relationships, and field definitions so AI tools can write accurate queries against real production databases — rather than inventing column names that don’t exist.
Industrial operations. Augury, an industrial AI company, released an OKF industrial profile in September 2026 — standard knowledge concepts for plant assets, sensor-trust policies, and operational relationships. AI agents across different factory systems now share a common vocabulary. Apache 2.0 licensed, fully open source. The same principle applies at any scale — from a factory floor to a ten-person SaaS company.
Where to Start
You don’t need to document everything. Pick one area where your business already struggles with consistency. Pricing rules. Support policies. Key metric definitions.
Write them down as plain text files with a simple YAML header. Put them in a folder. Add them to version control. That’s an OKF bundle.
No tooling required to get started. The tools — AI agents, analytics platforms, internal assistants — come later, when you want them to pull from that structured knowledge rather than making things up.
The companies getting the most from AI right now aren’t necessarily running the most powerful models. They’re the ones who’ve done the unglamorous work of organizing what they know. OKF is a standard way to do exactly that.
Want to explore how OKF could help your business build a better AI foundation? Let’s talk.