# Measure whether AI actually saved time

Original 5HO practice material. Fictional example. Reviewed 5 October 2026.

## Useful result

You'll finish with an honest comparison of elapsed time, quality and review effort.

A fast first draft is not automatically a time saving. Count the time spent checking and repairing it too. Otherwise the spreadsheet celebrates work you are still doing.

## Set up

Allow 15 minutes to set up a simple comparison. Use a timer, spreadsheet or calculator; AI can help organise the record but should not invent observations.

## Practice input

Manual run: 18 minutes. AI-assisted run: setup 6 minutes, prompting 3, review 5, corrections 2. Weekly frequency: 4. Only one run of each method has been measured.

## Do the task

1. Add all time spent on the AI-assisted run.
2. Compare equal-quality outputs, marking rejected results separately.
3. Repeat several times before projecting savings across a week or a team.

## Prompt

Calculate the observed time difference from this single trial. Show every component. Separate one-off setup from repeat work, but report the total run time as well. Do not claim a reliable average from one sample.

## Refine

Show the repeat-run estimate only as a scenario assuming the six-minute setup is reusable. Name that assumption next to the number.

## Expected checks

Total assisted time is 16 minutes, two minutes below the 18-minute manual run. Without setup it is 10 minutes, but that eight-minute difference is a scenario, not a measured repeat average.

## Save

Save the timing sheet and quality notes. Collect more samples before using the number in a business case.

## Sources

- Writing clear prompts: https://help.openai.com/en/articles/10032626-prompt-engineering-best-practices-for-chatgpt