Rethinking Design Workflows
One of my earliest trysts with AI experimentation was creating custom GPTs when OpenAI first launched the store. My initial exploration was nothing more than using AI to generate an image, rewriting copy, and so on. But after a brief phase of augmenting singular tasks, I started wondering if I could string several together to automate workflows, where the method took precedence over the tools.The question then became: where in the process does it actually make sense? Not every phase benefits from an AI layer - but discovery, ideation, research, client prep all have moments where speed or repetition is genuinely killing the work.Building something repeatable meant getting serious about structure. Prompting frameworks went from intuitive to intentional - from intructions to a structured system (architecture inspired by Aviad Cohen’s work with agentic systems). DEEP The first practical one I built. DEEP takes a company name and turns it into a comprehensive market research starting point in seconds - business context, competitors, consumer signals, market trends. Designed to require as little prompting as possible: all follow-ups run off a single index number from the menu. Zero to eighty, fast. IDE8 After discovery, the next bottleneck was ideation. IDE8 runs a How-Might-We question through multiple exercises of your choice - Idea Bingo, Quick Ideas, Crazy 8, Cheat Storm - turning a raw prompt into structured concept directions. Like having an extra workshop participant who never runs out of energy.User Reseacher Hearing from users is the one phase where nothing replaces the real thing. So rather than simulating user research, this simulates the interview - generating guides, running persona-based conversations, synthesising output into hypotheses. Not to replace it, but to prep for it. The head start before talking to actual humans.Commercial GPT Not every bottleneck is in the design process. I had a chance to jam with Manyone’s commercial director Bettina Lycke to design a GPT that could make prep for client meetings and negotiations more thorough. Commercial GPT simulates client conversations, discovery calls, negotiations, re-biz meetings, and gives live feedback on clarity, framing, and response handling. Built around frameworks like BANT. Useful before any high-stakes client meeting.