What I tried
Problem: I perform product discovery efforts on a twice weekly basis with a cross section of internal and customer end users. I communicate in process diagrams with these non-technical end users, but then have to quickly convert those visuals into text based requirements (user stories with Gherkin acceptance criteria) for development purposes. In my current organization structure, I am unable to take the full leap to full agentic development practices. I was looking for a hybrid-automation/manual pattern to experiment with without needing to overhaul existing development processes. Experiment: recalling a presentation I saw back in the 2000's related to the Prius's software being visual block diagramed and compiled into source code without human intervention (pre-LLM/Agents), I felt confident that I could devise instructions and steering documentation sufficient to take my block diagrams and convert them automatically into the user stories and format I required.
What happened
This has become my standard practice, removing ~1-2 days per sprint (5-10% productivity gain within a 3 week sprint cadence) and completely eliminating scenario gap errors (additional 5% gain based on development acceptance criteria change requests and UAT feedback).
Comments · 1
This is really cool seeing it visually! Are you creating the visuals manually still and then having it create the text or having both be generated?