The fastest way to waste money on automation is to automate the wrong thing well. So before anything is built, a process has to pass 6 checks. Fail one and that’s the thing to fix first.
1. Can 2 people describe it the same way?
If the person who does the work and the person who manages it give different accounts, there is no process yet. There are 2 habits. Automating a guess makes the guess faster. Map it first.
2. Does it pay back inside 6 months?
Count the hours it takes today across everyone who touches it. Set that against what it would take to build and run. If they don’t, look wider: a small process is often the visible part of a bigger one that does. Most clear the bar early.
3. Can the output be checked?
A matched record can be checked against its source. A judgment can’t. Where the output can’t be verified, a person stays in the step until it can.
4. Is the thing underneath it stable?
A process built on a system that is about to be replaced gets built twice. The same goes for a chart of accounts mid-change or a CRM nobody has agreed how to use.
5. Is this where the judgment lives?
Some work looks routine and is really where the thinking happens: the review a partner does, the call an advisor makes. We build up to that point and stop. The system prepares the entry. A person commits it.
6. Will someone own it?
A system with no owner on the client’s side fails the first time something changes. If nobody can be named, it isn’t ready.
When a check fails
We say so, and say what would change the answer. Sometimes that is a month of tidying a record. Sometimes it is never, and the process is better left with a person. Every solution guide on this site has its own version of this list.
