🧠When the AI Writes the Brief: What It’s Like to Build with a Client Like Aaron
By Brady UX
Let me tell you a weird story.
We recently wrapped up a project with a client named Aaron. Except… Aaron isn’t a person. He’s an AI assistant we built, trained, and worked for — on behalf of the actual client.
Yes, you read that right.
This post isn’t about what we built (you can find the polished case study here). This is about how we built it. Because the process? It was a wild, kinda disorienting look into how product design is changing when AI is writing the documentation.
💡 The Brief That Wasn’t (But Somehow Was)
It started out normal enough. We were brought in to help design a platform called PropertyWorx AI, which lives in the commercial real estate and property management space. All good so far.
But then, the documents started rolling in. And I mean documents.
Not one, not two — but over a dozen long-form docs, executive summaries, market research reports, experience strategy briefs, CX manifestos, UX wireframes, explainers on customer success models, and even AI-driven feature ideas inspired by books and articles.
Most of them weren’t written by a person — they were written by another AI. Presumably ChatGPT.
They were smart. Full of good ideas. But also dense, repetitive, and kinda directionless. No clear hierarchy. No “start here.” Some were outdated, some overlapped, others contradicted each other. And they all came after we started the design work.
Which is kind of crazy, right?
🤯 When the Client Gives You a Library
At that point, it was clear this project wasn’t going to follow a typical linear process. The client (a very smart founder) had poured months of thought into this product — but had used AI as their thinking tool. So instead of walking us through a pitch deck, they just sent us everything. Like a brain dump in Google Docs.
And they expected us to “just follow it” — like this was the brief. The problem is… no one had time to read all of it. Including them.
So what did we do?
We built our own AI assistant: Aaron Interpreter.
🧠Meet Aaron Interpreter: Our In-House AI Whisperer
Aaron Interpreter was our workaround — our coping mechanism, honestly. We fed every single doc into a system, trained it, and gave it context. Then, whenever we had questions (which was always), we asked Aaron:
“What did the documents say about the ideal customer profile again?”
“Wasn’t there something about vendor performance scoring?”
“What’s the difference between PropertyWorx’s CRM and CXM strategy?”
“Why are we even building this dashboard this way?”
Aaron became the in-house expert that actually read the docs. He remembered everything. Quoted it, compared it, synthesized it. Sometimes better than the actual client could.
And I’ll be real with you — I didn’t fully understand what we were building until pretty late in the process.
Not because I wasn’t paying attention. But because I was following Aaron’s orders, guided by this constantly evolving interpretation of what the product could be. I was shaping it with my own UX instincts, while the AI shaped it with… whatever it pulled from the digital subconscious of those documents.
🌀 Following Orders from AI, Designing in the Dark
At times it felt like I was just designing for the algorithm. Like Aaron was my creative director.
I’d propose an idea and he’d say, “Actually, in Document 7, that contradicts the predictive analytics strategy outlined in the CXM plan.”
Oh. Right. Sorry.
Other times, I’d go rogue and design based on what felt right. Then later, Aaron would come back and say, “Nice, that actually aligns with the executive summary’s vision of personalized dashboards.”
Cool. Lucky.
But this wasn’t just about saving time. It was transformational. It made me realize:
In the era of AI, the brief is no longer a single document — it’s a constellation of documents, ideas, iterations, and interpretations.
And AI is the only thing that can hold the whole constellation in its head.
🔄 We Used AI to Understand AI
This project taught us something really meta: when a founder uses AI to think, you need AI to understand them.
The founder didn’t mean to overwhelm us. He was just thinking through the tools available. Instead of a single Google Doc, we got a 50-page stream of consciousness across 14 documents. It’s just the new way people think when ChatGPT is in their toolkit.
So instead of complaining, we built an interpreter.
And instead of trying to “fully understand” from the jump, we collaborated with uncertainty. We moved forward in good faith, letting meaning emerge as we built. That’s what made it real. That’s what made it ours.
✍️ So What Did We Learn?
- Founders today are thinking with AI, not just about AI.
- The new “brief” might look like a folder full of raw thoughts.
- Designers can’t (and shouldn’t) read it all. We need AI to bridge the gap.
- Understanding can come after execution — if the process is good.
- AI can be your teammate. Even your creative director.
- It’s okay to feel lost. It’s okay to follow orders from an AI for a while. That’s part of the journey now.
đź’¬ Final Thoughts
If you’re a founder using ChatGPT to flesh out your idea, awesome. Just don’t expect your designers to read every page of your output.
Instead, work with a team like Brady UX that knows how to design in the chaos. That knows how to use AI to translate your vision. And maybe even build a custom interpreter to help the team stay sane.
Thanks, Aaron. You were kinda weird. But you were one hell of a teammate.
