SearchFree Newsletter

Atomic Scaling framework

JTO — Judgment, Taste and Override

Definition

JTO — Judgment, Taste and Override — is the Atomic Scaling framework naming the three decisions that stay human when AI agents run the work: judgment on ambiguous calls, taste on what is good enough to ship, and override on stopping an agent that is confidently wrong.

The problem it names: Automation quietly removes the human decisions a business actually depends on, and nobody notices until an agent ships something wrong.

When agents become the default operators of a company, the question stops beingwhat can we automate and becomes what must stay human. JTO is the answer in three words. It is the job description that survives automation — and, in anAI-native organisation, close to the whole of it.

Judgment

The call that cannot be specified in advance.

An agent hits an edge case at 2am — a client threatening to leave, a compliance flag with no clear rule. Judgment is deciding what should happen when the situation was never in the spec. It is the decision a company cannot delegate downward or outward, and the one automation is worst at.

Taste

The standard for what is good enough to ship.

Agents produce plausible work indefinitely. Taste is the judgment of whether plausible is good enough to put a name on — the quality bar, the brand voice, the thing a customer will feel. Without a human holding it, output volume rises and standards drift down.

Override

The authority to stop an agent that is confidently wrong.

Override is a named human with the standing power to halt, reverse or re-scope agent work — and the escalation triggers that pull them in before damage compounds. An organisation without an explicit override holder does not have human oversight; it has hope.

Why the acronym is not the point

JTO is a crowded set of three letters — it already belongs to a Joint Transportation Office, a Junior Telecom Officer exam and a Wikipedia disambiguation page. We use it as shorthand for people who already know the framework, and spell it out everywhere else. The idea is what travels: judgment, taste and override are the three decisions a company should never hand to an agent.

Next step

Where does human oversight sit in your operating model?

The Atomic Scaling Score reads your company across the six pillars of the 3P3R Method® and names the constraint holding growth back. Free, two minutes.

Take the Atomic Scaling Score →Talk to Ludovic about advisory

Questions people ask

What does JTO stand for?

JTO stands for Judgment, Taste and Override — the three decisions that stay human in an AI-native company, as defined by Atomic Scaling. It is unrelated to other uses of the acronym such as Joint Transportation Office or Junior Telecom Officer.

Who created the JTO framework?

JTO was named by Ludovic Bodin, author of Atomic Scaling and creator of the 3P3R Method®, as part of his work on Human-Assisted Autonomous Organizations.

How is JTO different from human-in-the-loop?

Human-in-the-loop describes a checkpoint in a workflow. JTO assigns three specific, named decision rights to people — judgment, taste and override — so oversight is a role someone holds rather than an approval step someone clicks through.

When should a human override an AI agent?

When the agent is operating outside the situation it was specified for, when the output would go out under your name below your quality bar, or when an escalation trigger fires. Atomic Scaling recommends defining those triggers before scaling past roughly two agents per person, not after.

How do I apply JTO in my company?

Write down, for each automated workflow, who holds judgment, who holds taste and who holds override, and what event pulls them in. If any of the three is unnamed, that is the gap. The Atomic Scaling Score shows where this sits against the rest of your operating model.