Where to start an automation

The usual reflex is to ask “what can we automate”. The right question is the reverse: what costs you the most time today, and needs no human judgement.

The rule that decides profitability

An automation is judged on two figures: the time it saves per month, and the time it demands to watch. A scenario that saves two hours and calls for one of maintenance is not a gain, it is a transfer of workload. We set those cases aside rather than sell them.

What automates well

Copying a form into a CRM. Creating a file when a client arrives. Sending a summary at a fixed time. Reconciling two lists that should say the same thing. Those tasks are repetitive, need no judgement, and cost hours in total.

What automates badly

Anything that assumes a decision. Sorting requests by importance, replying to an unhappy client, choosing between two priorities. An automation that takes those decisions produces errors nobody sees go by, until the day they cost a great deal.

The point we always frame

Who is told when a scenario fails, and how quickly. A silent automation that falls over on a Friday evening produces incomplete data all weekend, and nobody discovers it before the client reports it. Every automation touching a client or an invoice must raise an alert a human can see.