The Cost of Switching: An AI Lock-In Experiment (Post 1 of 6)
Can You Really Switch AI Tools?
For more than two years (starting in November of 2022), I’ve used ChatGPT almost every day. It became part of my workflow the way email or Teams is for most people, always open, always part of the process. So I started asking myself: what would it take to actually switch?
On paper, it looks easy. Claude and Google Gemini both cost $20/month, the same as ChatGPT. I figured I’d sign up, move over my context, and keep going. What I ran into was very different.
The first challenge was what I’ll call the illusion of portability. Sure, I could export my ChatGPT history, but Claude had no way to import it. Gemini could connect to my Google Drive, but the files I wanted to upload were too big. I found myself in command-line territory, splitting files into smaller chunks, debugging errors, and eventually making tradeoffs about what data was worth saving.
But even when I got the technical pieces sorted, the bigger cost became clear: cognitive switching. I had years of fluency with ChatGPT, I have my styles, my projects, my custom GPTs…. how to phrase prompts, what quirks to expect, which outputs I could trust. Starting over with Claude or Gemini felt like onboarding a new colleague who didn’t yet “get” how I work. Every task took longer, and every result needed more second-guessing.
That’s when it hit me: the $20 subscription isn’t the real lock-in. The real price is the accumulated trust, habits, and fluency you build with a tool over time. Research backs this up, 90% of workers now use AI, but 75% abandon tasks mid-stream because they don’t trust the results. source: https://www.innovativehumancapital.com/article/the-ai-at-work-adoption-gap-90-of-workers-use-it-most-still-don-t-trust-it
Once you finally learn how to “speak” to a system, it becomes very expensive mentally and procedurally to start fresh somewhere else.
This series will unpack that journey. I’ll explore portability, trust, choice, and organizational lock-in not just from theory, but through the lived experience of trying to switch.
So here’s the question I’ll leave you with after Post 1: if you lost all your AI history tomorrow, would you actually switch—or would you just rebuild with what you already know?
