Vic SolMeasure this before you automate: compared with what?
For anyone approving an automation who wants to defend the result a month later, with the "before" picture in hand.
When to use it
- The demo says the AI saves time and nobody has the figure for how long the process takes today.
- You are about to approve a pilot and want to know, in a month, whether it saved work or just moved it.
- A vendor is already reporting "savings" and you have nothing to argue with.
The sheet
The "before" picture. Taking it costs a week, not a project. Fill it before automating and fill it again one month later, with the same numbers.
| # | Number | Before (date: ____) | After, one month (date: ____) | How to take it |
|---|---|---|---|---|
| 1 | Volume | ____ per week | ____ per week | How many requests, documents or cases come in. |
| 2 | Time per unit, end to end | ____ | ____ | From the moment the request comes in until someone closes it. Time five real cases. |
| 3 | Error rate | ____ % | ____ % | Cases returned or disputed over the total. |
| 4 | Rework | ____ times | ____ times | How many times the same document gets corrected. Ask the person who does it, not the system. |
| 5 | Review time | ____ per unit | ____ per unit | The one nobody measures: how long goes into reviewing what came out, before and after the AI. |
Illustrative example: service quote · before: 20 per week, 3 days, 10%, 2 corrections, 15 minutes of review · after: 20, 1 day, 8%, 1 correction, 40 minutes of review.
Process: ____________ · Who does it today: ____________ · Who takes the "after" picture: ____________
How to fill it in 10 minutes
- Pick a single process, the one about to be automated. Do not measure the whole department.
- Fill volume from the system and the other four numbers from the person who does the work. The system does not know how many times a document was corrected.
- Write down the date of the "before" picture and book the "after" one for a month later, with a name attached.
- Leave row 5 for last and give it twice the time. It is the one that stays empty.
- The most common mistake: measuring only the time to do and forgetting the time to review. AI tends to move work from doing to reviewing; without measuring review, the savings get inflated.
The line that is almost always missing
Review time after the AI. In the illustrative example above, the process dropped from three days to one, and review rose from fifteen to forty minutes per unit. Without row 5, the report says "we saved two days". With row 5, the right question shows up on its own: how much of that saving went into reviewing? With the full "before" picture, the vendor no longer wins the argument for lack of data.
Vic Sol · Decision Desk · vicsol.org/resources/ · Free to use with attribution. This sheet structures a conversation; it is not legal, financial or audit advice.