Measure the part of the work you wanted to improve. If the aim was to reduce time preparing reports, measure that complete task. Customer retention is affected by too many other things to explain whether the drafting step helped.
The team’s observations are a useful starting point. “The reports are quicker” suggests timing the task. “They miss the important point” suggests checking whether each report answers the customer’s question.
Compare like with like
Here is a fictional comparison of similar monthly reports:
Measure | Before | During trial |
|---|---|---|
Reports completed | 12 | 16 |
Total preparation, checking and correction time | 12 hours | 12 hours |
Time per report | 60 minutes | 45 minutes |
Reports needing factual correction | 1 of 12 | 4 of 16 |
The total hours did not fall, but the team produced more reports. Time per report fell by fifteen minutes. The higher rate of factual correction needs attention, even if the errors were caught before delivery.
This example supports investigating the errors and the checks. It does not, by itself, establish that the tool caused either change. The reports may have been different or the people doing the work may have changed.
Count the work around the draft
Include finding information, preparing the input, checking, correcting and maintaining the instructions. If three people spend an hour discussing the tool, that uses three staff-hours.
Keep effort separate from turnaround. A report can take one hour to prepare and wait two days for approval. If the concern is late delivery, record that wait as well. Faster drafting may have little effect on the date the customer receives it.
Use comparable periods and work. A planned record before the change is helpful. If the tool is already in use, use available records or matched examples and explain any uncertainty in the comparison.
Read quality and time together
Check whether the same proportion of work was reviewed in both periods. Fewer recorded corrections can mean fewer errors, or less checking. Separate minor wording edits from errors that would change a decision or promise.
Include the team’s experience. Fewer interruptions may matter even where the time saving is difficult to measure. Describe that finding honestly without converting it into an invented percentage.
Use the evidence to decide whether to continue, change the approach or stop. If useful time has been freed, the saved-time guide addresses what you can do with it. Keep that choice separate from measuring how much time exists.
About Jay
I’m Jay Janes, a Practical AI Advisor and founder of Reformat Labs. I write about choosing useful work for AI, checking the results and supporting the people using it.