Open Knowledge Format (OKF): The Context Layer Agentic AI Has Been Waiting For

On 12 June 2026, Google Cloud published something most enterprise leaders haven't read yet — but probably should. The Open Knowledge Format (OKF) v0.1 is a vendor-neutral, open specification for packaging organisational knowledge — table schemas, metric definitions, runbooks, API documentation, internal processes — into a directory of plain Markdown files. No proprietary SDK. No new runtime. No lock-in. Released under Apache 2.0, it is available today on GitHub. Google itself describes it as "a starting point, not a finished standard."

It sounds like a technical footnote. I think it is a strategic signal — and I want to think through it out loud.

The Big Picture

A Simple Way to Frame Where We Are

For two decades, enterprise transformation has run in phases. We digitised data. We digitised process. What we never really digitised is organisational knowledge — the context that explains how the business actually works.

Most of Phase 3 still lives in the wrong places: definitions buried in Confluence, rules embedded in application code, and — most fragile of all — expertise sitting in the heads of a few senior people. That has always been a risk. Agentic AI turns it into a constraint.

The Data

And then the AI-specific pressure on top of those foundations: Gartner projects that 40% of enterprise applications will embed task-specific AI agents by end-2026, up from under 5% in 2025. The sobering counterweight, also from Gartner: over 40% of agentic AI projects are forecast to be cancelled by 2027 — with inadequate context and weak governance among the cited causes.

Read those last two together and you get the real story. Agents are arriving fast, and a large share of agent projects fail for lack of context. OKF is aimed squarely at that gap.

The Specification

What OKF Actually Does

The clearest way I can describe it: OKF answers the question "what does an AI agent know before it starts working?"

It sits upstream of the two patterns most enterprise teams already know. Think of it as the foundational layer:

Because it is just Markdown files in a git repo, every change to that knowledge is a commit — traceable, reviewable, revertible. For regulated industries, audit-readiness becomes a property of the knowledge layer itself rather than a bolt-on compliance exercise.

Three use cases feel immediately practical at enterprise scale:

• Data governance. A data team exports BigQuery table definitions, metric logic and join paths as an OKF bundle. Every change is a git commit — reviewable and revertible by default, without separate compliance tooling.
• Operational continuity. Each incident runbook becomes a structured document an on-call agent can traverse — without paging a senior engineer at 2 a.m.
• Cross-vendor portability. A vendor ships a catalogue export as OKF; your agents consume it directly, with no bespoke integration. The interoperability promise enterprise platforms have made for years, and rarely kept.

GCC & Product Leaders

Why This Matters Specifically for GCC and Product Leaders

Having spent years close to complex supply-chain and product platforms, this is where it gets real for me. An AI agent can read inventory tables and API schemas. But does it understand the difference between available and allocatable inventory? Why certain customers receive allocation priority? Which exceptions require human approval? That knowledge rarely lives in a database. It lives in people.

Codifying it into a governed format does three things that should interest any operating leader:

The Honest Question

The Question Worth Sitting With

OKF cleanly separates knowledge producers (people and pipelines) from knowledge consumers (agents). That separation is architecturally elegant. It also deserves a candid executive conversation: as we systematically externalise institutional knowledge into machine-readable form, what is our strategy for the people who built that knowledge?

I am not suggesting OKF is designed to reduce headcount — I genuinely do not believe that is the intent. But the history of enterprise technology is full of efficiency tools that reshaped workforce structures in ways nobody planned at launch. The honest answer is that the senior engineer spends less time re-explaining schemas and more time curating that knowledge — a higher-leverage role, if we choose to design it that way. That is a leadership decision, not a default.

OKF v0.1 is, by Google's own description, "a starting point, not a finished standard." The norms around knowledge ownership, workforce transition, and governance are still being written. And that is exactly why the conversation should start now, not after the rollout.

Sharing your experience in the comments would help more people than just me.