Judging honestly.
WHAT TO MEASURE
Time on specific tasks, before and after Output volume Quality, judged by whoever receives the work Errors reaching customers
THAT LAST ONE
Determines whether the time saving is real.
HOW TO MEASURE
Pick one task. Time it for two weeks before, two weeks after.
Actual comparison, not impressions.
WHAT USUALLY HAPPENS
Substantial saving on drafting Modest addition on verification Net positive for routine work Neutral or negative for work requiring precision
WHAT TO WATCH FOR
Volume rising while quality falls Verification skipped as familiarity grows Customer complaints about generic communication
WHEN TO REVIEW
Quarterly in the first year, annually afterwards.
WHAT TO DO WITH POOR RESULTS
Establish whether it is the tool, the task, or how it is being used.
Usually the last. Poor prompting produces poor output.
THE QUESTION THAT MATTERS
Would you go back to doing it without? If yes, stop.
