Brief — daily briefing for ad accounts

Founding Issue

An advertising decision layer for people who already know how to run ads.

Most ad tooling tells you what already happened. Market Agent tells you what to do about it — in the language a senior UA manager would use, with the numbers backing each call, and a single click to accept or reject.

The model reads the same data your team does — Google Ads today, GA4 next — and writes a short case file each morning: what changed, why it matters, what to do, what could go wrong. Nothing executes. You decide. The disagreements become the spec.

Enter the editionInvitation-only · POC · v0.1

Data sourcesGoogle Ads → GA4

Output3–5 case files / day

MutationsZero

ApprovalAlways human

The thesis, in three notes

Read-only by hard rule.

No mutate, create, update, or remove calls anywhere in the ingestion path. The model proposes; a human disposes. We treat the absence of a write path as architecture, not a feature flag.

II.

The agent reads. We inject the expertise.

Beneath each case file is a stack of specialized agents and assisted ML models that codify what senior UA managers already know — the failure modes, the gut checks, the patterns Google’s native recommendations don’t surface. The goal: sharpen the decisions you’d already be making, and let you react to reality faster than you can on your own.

III.

Foundations before flourishes.

Numbers in our canonical metric layer must match the Google Ads UI side-by-side before a single recommendation is generated. If the foundation is wrong, every reading is built on noise.