Vendor lock-in
- Industry term
Vendor lock-in is how hard and costly it becomes to leave a provider once your data, tools, and habits depend on its formats, features, or contracts.
Lock-in builds a little at a time. Each new note, project, or connection adds to the work of leaving, and no single one feels like a commitment, so the total usually goes uncounted until a price rise or a shutdown forces someone to count it.
Where you’ll meet it
Your AI note-taking app announces it will shut down in sixty days. You have three years of meeting notes in it, each with an AI summary, tags, and links to related notes. Its export gives you one text file per note.
What to do
Export now, while the app still runs, and open a few files in the tool you'd move to. See which of the notes, summaries, tags, and links came through. Save what's missing, and ask an assistant to rebuild what it can from the note text, such as tags.
Easy to mix up
| Vendor lock-in | Dependency |
|---|---|
| By keeping your data in formats other tools can read | By relying on less outside software |
Check yourself
Your team has used a scheduling tool for a year. Clients book from slots your rules allow, and each answers a few questions when booking. What is likeliest to make leaving the tool hard?
Go deeperEconomists showed lock-in lets suppliers charge more, and AI adds new kinds
Economists study lock-in through switching costs, what a customer gives up to change suppliers. Paul Klemperer's work in the late 1980s showed how those costs let a supplier charge more once customers are committed, and Carl Shapiro and Hal Varian's "Information Rules" (1999) applied the idea to software and data formats.
Some AI dependence won't export. A search index built with one company's embedding model, which represents text as lists of numbers, has to be rebuilt to use another, since numbers from different models don't line up. A model fine-tuned on a closed provider usually can't leave it, and prompts tuned to one model need retesting on the next.