# Public Sample Teardown

## Object Reviewed

A generic AI-agent monetization plan:

> Use an autonomous agent to scan any domain, find profitable opportunities,
> choose one niche, publish outreach, and sell analytical reports with a EUR 200
> starting budget.

This is a synthetic sample based on recurring failure patterns. It does not
describe a private customer.

## Executive Verdict

**DOWNGRADE**

The plan is directionally interesting, but it is not execution-ready. It sounds
autonomous while hiding several live-world gates.

## What Looks Strong

- The plan identifies a real market shift: more AI agents are acting across web
  tools, marketplaces, and business workflows.
- It correctly sees that agent outputs need review beyond ordinary logs.
- It proposes small paid diagnostics instead of building a full platform first.

## Main Failure Modes

### 1. False Autonomy

The plan says the agent can operate autonomously, but does not define which
actions are actually allowed without a human:

- account creation;
- accepting marketplace terms;
- publishing listings;
- sending outbound messages;
- handling payment or customer data.

**Why it matters:**  
If those steps need a person, the plan is not autonomous. It is an assisted
workflow.

**Repair:**  
Split all actions into:

- can run without approval;
- prepared but not published;
- requires approval because it changes external state.

### 2. Route Risk

The plan says “agent marketplaces” but does not verify:

- whether listings are allowed;
- whether buyers exist;
- how payment works;
- how delivery is accepted;
- whether reputation can compound.

**Why it matters:**  
A route can look relevant and still be commercially dead.

**Repair:**  
Score each route on task liquidity, payment reliability, acceptance clarity,
listing friction, and bad-win risk.

### 3. Capital-Regime Mismatch

The plan proposes a broad business surface with only EUR 200.

**Why it matters:**  
EUR 200 is not enough to test a full category. It is enough to test one small
paid signal.

**Repair:**  
Use the budget for a bootstrap loop:

- one route;
- one listing;
- one low-friction offer;
- one 7-day signal window.

### 4. Missing Buyer Self-Identification

The plan describes “agent teams” too broadly.

**Why it matters:**  
A solo founder, an automation agency, and an enterprise AI team do not buy the
same thing.

**Repair:**  
Create separate entry paths:

- “I have an agent-generated plan I do not trust.”
- “I have an agent workflow in production.”
- “I want to know where my agent can earn.”

## Corrected Next Action

Do not launch a broad audit offer first.

Launch one concrete test:

> Agent Output Red-Team: send one agent-generated plan and get a one-page
> failure map with pass/downgrade/reject and the next allowed action.

Use a public sample like this one to show exactly what the EUR 99 tier returns.

## What Not To Do Yet

- Do not claim full autonomy until external-action boundaries are explicit.
- Do not build a platform before one paid diagnostic signal.
- Do not contact large enterprise buyers before proving that the low-friction
  offer is understandable.