Define what finished looks like
Choose one task with a clear start and finish. For a weekly program update, the start might be opening the latest activity notes and the finish might be an approved summary ready to send. Ask the people who do it to track how much time they spend on preparation, writing, review, and corrections. Note time spent waiting separately from the time staff are actively doing the work.
Record how often the task happens and how many people take part. If two reviewers each spend time on it, count both. The Less Desk Time Ledger can help estimate the weekly total. Save the assumptions behind the estimate so you can make a fair comparison later.
Measure the new process the same way
Try the same task with the process that uses AI. Include time spent preparing the inputs, checking the output, correcting errors, and getting a final approval. Track setup and training time separately. Keep ongoing maintenance in view, including time someone spends updating instructions or fixing the workflow when a source changes.
An illustrative example is a report that used to take 60 minutes of staff time. If staff now spend 10 minutes preparing the input and 25 minutes reviewing and correcting the draft, the net saving is 25 minutes for that run. Repeat the comparison with a routine report and one with missing information. Averages can hide the days that need the most attention.
Check the quality and what happens to the time
Agree in advance on a few signs that the output is good enough. Does the report use the right figures? Does it include the context a reader needs? Is someone clearly responsible for the final version? A faster draft that creates more corrections for another team may simply move the work.
Ask staff what they could do with the time they get back. They might be able to call partners sooner, spend more time with a client, or catch up on something that always gets postponed. Compare the time, the quality, and the new capacity before expanding the automation to another task.