How I Set Up a System of Custom GPTs to Scale Creative Direction
As a lead visual designer, my job isn’t just to make things look good—it’s to steer strategy, storytelling, visual systems, and stakeholder alignment, often all at once. When the projects stack up and timelines collapse (as they often do), context-switching becomes a productivity killer. I found myself needing something unusual: a way to scale my brain.
So, I built a system of custom GPTs, each assigned to a distinct creative role. Together, they act like an AI-powered team that helps me think through strategy, write copy, develop visual metaphors, build design systems, and even prepare stakeholder walkthroughs.
Here’s how I designed the system, how it actually works behind the scenes, and how it's helping me lead smarter and faster across high-impact creative projects.
The Spark: A Team of Digital Collaborators
It started as an experiment: could I train one GPT to help with strategy? Then another to assist with copywriting? Once I realized I could chain multiple custom GPTs together, each with its own personality, guardrails, and role, I started treating them like my creative team.
I mapped out the typical phases of a project and built a custom GPT for each one. Sidenote: I built a custom gpt to help me build custom gpts because why not.
The Cast of Characters
Each GPT is a specialist. Here's the current lineup:
StrategistGPT – Shapes the high-level concept, tension, and narrative arc
VoiceGPT – Crafts core messaging pillars, headlines, and subheaders
VisualizerGPT – Develops visual metaphors, image prompts, and palette ideas
SystemsGPT – Translates visuals into scalable design tokens and layout logic
FixerGPT – QA for contrast, accessibility, spacing, and responsiveness
ProducerGPT – Builds rollout plans, task lists, and timelines
CitationGPT – Verifies legal language and data accuracy
DemoGPT – Writes polished scripts for stakeholder walkthroughs or pitch decks
Each GPT is trained with voice, context, and design standards that are specific to my work so they plug directly into my workflow.
How It Actually Works
It’s a structured system that lives inside my ChatGPT account, optimized for how I lead creative projects.
1. Each Project Has Its Own Folder & Instruction Hub
Every project I’m working on has its own folder inside ChatGPT. In each one, I’ve pinned an Instruction Hub—a doc that contains:
A brief overview of the project
A clearly defined order of operations (e.g., Strategist → Voice → Visualizer → Systems...)
Direct links to each custom GPT I built, with notes on when and how to use them
Because it’s all internal, my team doesn’t need access to the ChatGPT backend. It’s designed to support me as the creative lead—streamlining the invisible labor that keeps everything running smoothly.
2. Swapping in Humans Where It Makes Sense
If I’m collaborating with a human (like a strategist, writer, or illustrator), I can simply note that in the Instruction Hub and skip the GPT for that role.
What’s powerful is that the system still works. I use the GPT output or structure to:
Define the handoff more clearly
Give collaborators a strategic starting point
Optimize our collaboration by removing ambiguity
This is what has saved me the most time: I’m no longer waiting on a comms director or copywriter to provide a full comms plan and copy options from scratch. I generate a solid first draft myself using my custom gpts, then share it with them—making it clear that it’s a starting point, not the final say. They’re free to rewrite, reframe, or challenge it. But they’re not starting from a blank page, and that makes our work faster, smoother, and more collaborative.
It’s not about replacing people—it’s about creating a flexible structure that supports both AI- and human-powered creativity.
3. GPTs Work in Sequence
Each GPT contributes one building block in the creative process. The order matters:
StrategistGPT sets the creative vision
VoiceGPT transforms it into messaging
VisualizerGPT builds the visual concept
SystemsGPT scales that into reusable assets
FixerGPT QA-checks the work
ProducerGPT turns everything into a launch plan
CitationGPT checks for legal/data issues
DemoGPT prepares the final walkthrough script
It’s like a production line—but for ideas. Each role feeds the next.
4. It’s a Living System
If something changes, I tweak it. I might refine the prompts inside one GPT, change the order of operations, or add a new GPT altogether—like AVGPT, which oversees audio/visual planning: everything from stage specs to lighting cues and screen states for live events.
I’ve never worked closely with AV teams before, so I built AVGPT to help me learn what to ask, what to clarify, and how to better anticipate their needs. It’s making my communication more informed and our planning process way more efficient. As I gain insight into what our auditoriums are capable of, I’m adding those specs and notes into my custom AVGPT instructions, so it improves on itself.
Because the system is modular, it’s easy to adapt as I grow. Each new GPT is a way to close a knowledge gap or smooth out a bottleneck—without adding overhead.
A Quick Note on Confidential Work
While this system is incredibly helpful for scaling creative direction for our employee events, I never use it for confidential projects. Anything involving proprietary data, unreleased IP, or sensitive content stays completely outside this AI system. Technology pitches, forecasting, business strategy…I keep my operations old school for those.
The GPTs support ideation and execution—but discretion still comes first.
What’s Next
Eventually, I’d love to integrate this system with project tools like Relay.app or Notion to create a fully automated command center—one where project briefs, timelines, and GPT outputs are all seamlessly connected.
And while this system started as a personal productivity tool, I see its value growing far beyond just my own workflow. I’m building an internal AI collaboration program at Walmart Global Tech, designed to help others explore what’s possible when you pair human creativity with smart agents.
After combing through internal Slack channels and AI forums, I’ve identified a group of “AI Champions”—colleagues who are already building custom agents or experimenting with AI in powerful ways. So far, more than 30 developers and engineers have volunteered to share how they’re using AI to supercharge their work.
The goal? To bring the rest of the org along. I’m building a spotlight series that demystifies AI workflows and showcases real examples—especially for associates who aren’t technical, but who want to learn how to collaborate with AI in ways that are approachable, empowering, and grounded in actual business use cases.
Because the more minds at the table, the more creative (and useful) this future becomes.