Scene Hook
Last Thursday at 1 AM, a client urgently needed a wording change in an automated inquiry reply. I opened the email routing logic ChatGPT helped me build three months ago, staring at a dozen connected modules on the screen, completely clueless about which line did what. At that moment, I realized: the three hours I saved using AI might cost me thirty hours to pay back.
What It Is + Who Is Using
Margaret Storey recently wrote an article calling this phenomenon "Cognitive Debt." Unlike traditional technical debt—where we know what's poorly written—cognitive debt means we have no clue what we actually built. I messed this up before: last year, I used AI to build five automation workflows in one breath without documenting anything. Three months later, I needed to tweak a filter condition, spent two hours hunting for where to make the change, and ultimately had to delete and start over. Freelance illustrator Zhang Wei fell into the same trap. She used ChatGPT + Make.com to build a client inquiry routing system. Last month, when she needed to add a new rule, she opened the backend, stared blankly at the tangled wires, and eventually tore it all down to rebuild, wasting an entire afternoon. This isn't an isolated case. Among over 60 comments on Hacker News, many indie hackers and one-person companies report the same issue—AI lets us run faster, but we understand less about what we've created.
Replicate Cost
Reducing cognitive debt doesn't require buying new tools, just one habit: every time we build something with AI, spend 15 minutes writing a "what is this" explainer. Money: $0. Time: 15 extra minutes per automation workflow for documentation. Technical barrier: If we can type, we can do it; no coding skills needed. First step: Create a new document, title it with the workflow name and build date, then write down three things in plain language—what the workflow "does," "what the input and output are," and "where the key decision logic happens."
Advice By Stage
For those just starting out without AI automations yet—I recommend creating a "My Automations List" document now, building the documentation habit from the very first workflow so things get easier later. For those with 1-2 clients already building with AI—I suggest spending 30 minutes today going back to add an explainer for each workflow, to avoid getting stuck at 1 AM like I did. For those scaling with a team using AI—we should set a rule: whoever builds it writes the explainer, and a workflow without one isn't considered done. Also, this isn't for everyone; if operations are entirely manual right now, there's no rush. When starting to build in bulk with AI, coming back to pick up this habit then works fine too.