Before
Record how long the work takes today and how often it needs correcting.
Current work
After
Count time spent checking and correcting the new result, plus its cost.
First test
Decide
Compare the result with what you agreed at the start.
Keep, change or stop
Set up the comparison
- Name the task, the person responsible, the reviewer and the review date.
- Choose similar periods before and after the change. Record how many examples you checked and how varied they were.
- Write down what counts as a good result and what you hope to improve before the first test.
- Keep the source of each number. Note changes in staff, demand or policy that could affect the comparison.
| Measure | How to record it |
|---|---|
| Time per accepted result | Minutes spent preparing, checking and correcting / results accepted. Use similar tasks in each period and include the time a person spends checking. |
| Results accepted | Results accepted / results checked x 100. Agree what counts as accepted before collecting results. |
| Results needing correction | Results corrected / results checked x 100. Count each result once, even if it needs several changes. |
| Task result | Choose one result that matters for this task and say exactly how you count it. Note other changes such as staffing or seasonal demand. |
| Ongoing cost | AI service + hosting + other direct usage cost during the period. State the currency and keep setup fees separate. |
| Problems needing attention | Number of cases that needed help or correction. Record how serious each problem was and how it was resolved. |
A worked example
Example: the current method takes 240 minutes for 8 accepted reports, or 30 minutes per report. The new method takes 180 minutes, including review and corrections, for 9 accepted reports, or 20 minutes per report. That is 10 minutes less per accepted report. These made-up figures are not Genaima customer results.
Do not compare total minutes alone when the number of accepted reports changes. If none are accepted in a period, record that failure instead of calculating time per report. Less time spent does not automatically mean money saved; you would need to show an actual change in cost or available staff time.
Use the worksheet
Download the CSV and open it in your spreadsheet tool. Fill the period, value, source, owner and target columns for each measure. The template contains definitions and empty measurement cells; it does not calculate results for you. Keep implementation fees separate from recurring usage costs, and retain the underlying logs.
Make the review decision
Compare the result with the goal agreed at the start. Check rejected cases, mistakes and the work needed to keep the task running. Decide whether to continue, change it or stop, and name who will follow up. Faster drafts are not useful if quality falls or checking takes too long.