Measuring automation ROI before you build anything

Before building an automation you should be able to say roughly what it is worth, and after building it you should be able to check. Both need the same thing: a baseline measured on the real process.

Start with the baseline

Measure four things on the process as it runs today:

• Time per item: how long one unit of work takes the person doing it.
• Volume: how many items arrive in a normal week.
• Rework: how often an item has to be redone or corrected.
• Waiting time: how long items sit before anyone touches them, which customers often notice more than the work itself.

Time about twenty real items with the people who do them, and note the spread. Averages hide the difficult cases.

A simple estimate

Here is an illustration with made-up numbers. Suppose a task takes 6 minutes, happens 40 times a day, on 5 days a week. That is 6 x 40 x 5 = 1,200 minutes, or 20 hours, every week.

Now suppose an agent prepares each item and a person reviews the result in about a minute. Review time is 1 x 40 x 5 = 200 minutes, a little over 3 hours a week. The difference is 1,000 minutes, roughly 16 to 17 hours a week, before counting the cost of building and running the agent. Your numbers will differ, which is exactly why you measure.

What the estimate leaves out

• The cost to build, run and maintain the agent.
• Time spent on exceptions the agent cannot handle.
• Quality: fewer errors, fewer corrections and less rework.
• Speed: answering a customer or closing the month sooner.
• Risk: fewer things that depend on one person remembering.

Measure again afterwards

Use the same measures, the same method and, if you can, the same people. Compare against the baseline, not against memory. Then decide honestly whether the result is worth widening.