Who owns the judgment.

The question of the AI era is not a technology question. It is a governance question: which decisions stay human, and how that line holds in writing when the machines get good.

Our doctrine is the 80/20. Agents industrialize the repeatable eighty percent of the growth operation — bid operations, feed maintenance, creative rotation, QA, reporting — work where the procedure is known and the volume is the challenge. Senior operators own the twenty percent where judgment compounds: where the money goes, what a brand is willing to say, and when something stops.

The split is decided by design, before anything is built. Every task in the operation is routed — not by what a model can plausibly do, but by what the P&L can afford to have done without a person in the loop.

The line holds in writing

Three mechanisms keep the split honest. Approval thresholds define what each agent may spend and change on its own; above the line, a person approves, and the thresholds are set with the client, in writing. Kill criteria are published before launch — every automation ships with the conditions that will retire it, agreed in advance. And metered cost prices every agent action into fully loaded CAC, tracked against the human hour it replaces.

An automation earns its seat the way an operator does: outperform the baseline, or be retired.

The consequence is symmetry. An automation is held to the same standard as a hire — measured output against the human baseline, its costs on the books, its termination conditions signed before its first day. When performance drops below the baseline, it is switched off. No sunk-cost defense, no pilot that never ends.

Leverage is judgment, concentrated

Run this way, the org chart changes shape. A few senior operators run what took twenty people — and the leverage comes from concentration, with each operator’s judgment applied across the volume the agent layer creates. The scarce resource in the AI era was never machine capacity. It is the person who knows what to do with it — and an operating model that keeps that person accountable for every dollar the machines touch.

← All Pareto Insights