Who is your AI's knowledge for?

A few days ago, Google Cloud published the Open Knowledge Format (OKF): an open standard for the “knowledge” AI agents read to understand a company’s systems and data. For us it was a strange kind of good news — it describes, almost part for part, the shape we’d already been building for months. When a company with Google’s reach independently arrives at the same design, the bet looks right.

But the validation isn’t the interesting part. The interesting part is a choice the standard quietly puts in front of every company — and it has nothing to do with file formats.

Two roads for the same knowledge

Ask one question about any knowledge your team keeps for its AI: who is it for?

One answer — the one most of the industry is racing toward — is the agent itself. You write down everything the AI needs so it can run on its own, longer, with people out of the loop. The dream is an agent that takes a one-line instruction and works for days unattended. OKF is good infrastructure for that world: it makes your data legible to the machine.

The other answer is your people. You keep that knowledge so an engineer, a manager, or an auditor can see what was decided and why, what’s underway, what’s still pending, and where the work is heading — without having to read every line of code. The AI still does the heavy lifting; the difference is that a human stays oriented and in command.

The question is no longer how capable the AI is. It’s whether the people responsible for the result can still see what it’s doing.

Why the second road is the one that builds trust

We’ve written before that the real obstacle to AI at work isn’t capability — it’s trust, and trust is built with consent, traceability, and accountability. An autonomous agent you can’t see into is a liability waiting to surface in an audit, a handover, or an outage. Knowledge that keeps a person in the loop is the opposite: an asset that makes the same AI safe to rely on.

This is the whole reason StrayMark, our open-source tool, takes the path it does. It doesn’t exist to make you use more AI. It exists so the AI you already use leaves a record — and a map — that a human can actually stand inside.

A standard like OKF can hand you the envelope: the format, the plumbing, soon a commodity that benefits everyone. What it deliberately can’t decide is who your knowledge serves. That part was always yours to choose.

(For the technically inclined, we wrote the detailed version on the StrayMark blog.)