Derek
Murray
Design Director + Creative Lead · Brand, Design Systems & Executive Communications
DM · ORLANDO, FL
Fast-growing companies hit the same wall: the brand that looked sharp at launch starts fraying the moment more hands touch it. I fix that. For 15+ years I've built identities from zero, built the teams around them, and built the systems that keep every touchpoint on-brand without me in the room. My bar: work that reads in three seconds and earns its keep.
Five master templates in, 700+ finished ad variations out. The run took three designers six hours a day for a week. It now takes one designer an hour.
Chained so a brief goes in and production-ready assets come out, with a named reviewer signing every frame before release.
An agent that holds brand rules, voice and templates, so designers ship on-brand without routing every decision through me.
One template system localized across 100+ countries, plus a self-serve portal that cut a two-day request queue to a download.
The challenge. Generative AI dramatically lowered the barrier to creating content, and introduced a new problem for creative organizations. Individual tools could produce impressive images, copy, concepts, and variations, but isolated tools do not make a scalable creative workflow. The real question was never which model is best. It was: how does a business brief become an AI-ready creative brief? How do strategy and brand standards survive across multiple models? Which model handles each part of the process? What happens when a model returns something unusable, and where does human judgment enter? The answer was to stop treating Generative AI as a set of tools and start treating it as creative infrastructure.
Turns the business objective into structured creative direction: objective, audience, positioning, brand rules, channel specs. The output isn't a prompt; it's a creative specification downstream models execute against.
Decides what happens next and which model performs each task, using model chaining, APIs, agent frameworks, and node-based environments. Separates the workflow from the model, so components stay interchangeable.
Concepts become assets. The right engine is chosen per brief rather than forcing every project through one: controlled production imagery, concept exploration, conversational editing, typographic work, illustration and icon systems.
Stop regenerating; start directing. Inpainting, outpainting, image-to-image, object and background replacement, canvas extension, upscaling. An approved asset becomes a master: 1:1 extends to 16:9 and 9:16 and a full campaign family.
Control returns to human creatives. Typography, grid, identity, color accuracy, compositing, retouching, accessibility, channel specs, creative QA, answering the only question that matters: would a professional team actually ship this?
The principle. No single AI model should be responsible for the entire creative process. Each layer solves a different problem, and because the layers are separate, a better image generator six months from now is a component swap, not a rebuild. The models are commodities. The workflow architecture is not.
Diagnose the layer, not the prompt. When something fails, the failure has an address. That distinction is what stops teams from trying to solve every problem by prompting again.
The strategic shift. Most organizations are still in the middle column.
The outcome. That final loop is the real opportunity. As campaign performance, creative decisions, model behavior, brand preferences and human feedback become inputs to future workflows, the production system becomes increasingly specific to the organization using it, and increasingly hard to copy. The five-layer architecture scales Generative AI while preserving the things that make professional creative work valuable: strategy, taste, brand consistency, creative judgment, production discipline. The goal was never to remove creatives from the process. It was to remove production friction so they spend their time on the decisions machines are least qualified to make. Generative AI creates assets. A production system creates creative leverage.
The case study above, expanded into a field guide for creative leaders. The five layer architecture in full, plus the parts that only show up once you run it: the four gates that decide whether a generated image is usable, the diagnostic table that stops your team regenerating a strategy problem, two worked pipelines, and a one page contract template.
Identity and design system from zero, made for the people who'll use it.
Hire, train, and set review rhythms so quality holds without a gatekeeper.
Put AI into production so design hours go back to craft.
Your note is on its way to derek@wecandothat.co. I read every one and reply personally.