Adstra Standard Operating Procedure: AI Ad Creation & Lifecycle Management

Video going over it

This document outlines the strict, end-to-end workflow for creating, uploading, and managing Meta ads for Adstra and its clients. This process leverages AI (Claude and Nano Banana) to generate high-converting creatives while maintaining strict adherence to Adstra's media buying architecture and Notion tracking systems.

1. Project Context & Setup

To maintain historical context, research briefs, and performance data, every client must have their own dedicated Manus project/thread.

Do not mix multiple clients in a single chat. If starting work for a new client, create a new project named [Client Name] Ads Project. This ensures Manus retains the specific context of what is working for that client.

2. The 3-Stage Creation Workflow

The ad creation process is divided into three distinct stages. Manus will execute these stages sequentially and will ask for your approval and required variables before proceeding.

Stage 1: Prompt Generation (via Claude)

Before generating any images, Manus will use Claude to engineer strategic image prompts based on the client brief.

Required Inputs:

The 60-80% Rule:

60% to 80% of all new ad concepts generated must be based solely on the winning ad template. These variations should feature minimal changes, as the core structure is already proven to convert. The remaining 20-40% can be experimental concepts.

Manus will present the generated prompts (including the powerful headline and relevant CTA) for your review. You must approve these prompts before moving to Stage 2.

Stage 2: Image Generation (via Nano Banana)

Once the prompts are approved, Manus will generate the actual ad creatives.

Creative Requirements: