Test 1: does it take more than three hours a week?

Count the hours honestly, over a normal week. Below three hours a Sprint pays for itself slowly. A standard tool, or simply a better agreement, is then the better choice.

Test 2: are there rules?

A quote from a price list has rules. A creative proposal for a new client does not. AI can cope with variation in the input, but not with the absence of a decision rule. If you can write down in ten minutes when the answer is right, it can be done.

Test 3: is there a person for the exceptions?

Every automation produces exceptions. Someone has to be willing to review them, every day, in a few minutes. If that person is not there, the exceptions pile up and the team goes back to manual work.

Test 4: can the data come out of the system?

Does your software have an API or an export? Exact, Moneybird, AFAS, HubSpot and most modern tools do. An old custom package without an integration often does not. Then it gets expensive, and we say so up front.

Test 5: does a standard tool already do it?

For many tasks a tool exists at EUR 30 a month. Building your own only makes sense if you spend more than EUR 500 a month on separate tools for the same process, or if the tool does not follow your way of working.

What we do when a test fails

Then it goes in the Quickscan report, with the alternative. A no costs us an assignment and gives you a trusted address for your next question. That is a good trade.

Last updated: 2026-10-05.