Back to technical studies
TECHNICAL STUDY

Designer Agency: When a Creative GPT Needed an Operating System

A creative-agency GPT became an experiment in coherent instructions, handoffs, and honest reviews.

A creative agency inside a GPT sounds simple: a few specialist roles, a design brief and a list of deliverables. But what happens when the brief changes or one specialist contradicts another? Designer Agency / Agency MyGPT became my way of investigating that question.

From a team of names to a working method

The earlier Agency package was built around creative deliverables, visual direction and prompt libraries. By late January 2026, I was comparing it with a more reusable MyGPT system. My question in the archive was practical: had the merger actually accomplished anything?

The review documented a clearer main instruction, execution flow, governance, guardrails and operations playbook. That's an architectural improvement in the reviewed material, not independent proof that a group of custom GPTs was delivering client work.

What got better, and what got thinner?

Consolidating instructions can reduce contradictions. It can also flatten the specialist depth that made the original creative library useful. Later, another review could see an Agency ZIP's name without being able to open its internals. A confident description was no substitute for accessible evidence.

The useful return from this experiment was a stronger question: can I trace a creative decision from the brief to the specialist, the delivered artifact and its review? That is worth more than a growing list of impressive-sounding job titles.

This is an editorial account of documented package design and reviews, not a claim of measured customer productivity or a production agency service.

All writing