A manager can reach an understandable conclusion: if AI improves productivity, adoption should be accelerated. Require teams to use a certain tool, for example, or set a minimum number of prompts.
The problem starts when use of the tool becomes the goal instead of the result.
People optimize the metric
If prompt volume is measured, prompts will appear. If the share of employees using the tool is measured, active accounts will appear.
Neither proves that work became faster, cheaper, or better.
AI is not useful for every task
There is work where AI saves hours. There are tasks that are faster to do directly. And there are processes where reviewing AI output consumes the entire gain.
A single rule ignores differences between roles, risks, and types of work.
Mandatory use can reduce quality
If employees feel they must demonstrate AI use, they will be tempted to insert it where it is not needed.
The organization may produce more intermediate text, code, or analysis without producing more useful outcomes.
Leadership should remove barriers instead
Rather than ordering people to use AI, provide safe tools, clear data rules, training, and room for experimentation.
Then the strongest use cases can become standard practices where they actually work.
Measure how the work changes
If AI is introduced into a process, look at cycle time, quality, cost, manual steps, and customer outcomes.
Usage can be a supporting adoption metric, but it is not evidence of value.
Expecting efficiency is reasonable
A company can absolutely expect employees to use modern tools and search for more effective ways of working.
But it is stronger to define the goal in terms of outcomes and keep the tool as a means unless there is a specific reason to standardize it.
If AI genuinely makes the work better, employees will not need much persuasion to use it. If the company needs a KPI for prompt volume, it may be measuring the wrong thing.