Standard AI adoption surveys overwhelmingly measure uptake: what share of employees or organizations have started using a given AI tool. A structurally similar and arguably more informative question is almost never asked in the same surveys: what share of those who started have since quietly abandoned the tool, and why.
Why abandonment data is so rarely collected
Uptake is a simpler, one-time measurement that fits neatly into a survey snapshot. Abandonment requires tracking the same population over time and specifically following up with people who've stopped using something, which is more methodologically demanding than a single point-in-time adoption survey and correspondingly rarer in practice.
Why this gap distorts the overall adoption picture
A survey measuring only cumulative uptake - anyone who ever tried the tool - can show steadily rising adoption even in a population where a large and growing share of early adopters have actually abandoned the tool, because the survey instrument simply isn't designed to detect abandonment in the first place.
What abandonment data would reveal
The specific reasons employees give for abandoning an AI tool - too unreliable for the task, too disruptive to an existing workflow, insufficient time savings to justify the learning curve - are exactly the kind of information that would help organizations understand where AI tools are and aren't delivering real value, information that pure uptake numbers can't provide.
What better adoption research would look like
Adoption surveys that specifically track and report on abandonment alongside uptake, treating the two as equally important parts of the same picture, would give a far more honest and useful account of how AI tools actually perform once the initial novelty wears off - a more complete picture than the uptake-only numbers that currently dominate this kind of research.