A simple way to decide what to automate first
Small teams do not lack automation ideas, they lack a way to rank them. A one-page scoring method we use in consulting engagements.
Every audit we run ends with a long list of things that could be automated: invoice matching, onboarding emails, report assembly, data entry between two tools that refuse to talk to each other. The list is never the problem. Deciding what to do first is.
Score four things
For each candidate, estimate four numbers. Rough is fine; the goal is ranking, not accounting.
- Hours per month spent on the task today, across everyone who touches it.
- Error cost: what a mistake costs when it happens, times how often it happens. Late payments, angry customers, rework.
- Build effort in days, including the boring parts: access to systems, edge cases, monitoring.
- Fragility from 1 to 5: how likely the process is to change next quarter. Automating something that will be redesigned in two months is waste.
Rank by payback
Monthly value is hours saved times a loaded hourly rate, plus the error cost avoided. Divide build effort (in money) by monthly value and you have payback in months. Sort ascending, then push anything with fragility 4 or 5 down the list regardless of payback.
What usually wins
In our engagements the top of the list is rarely the flashy AI idea. It is the weekly report someone assembles by hand from three exports, or the customer email that a human copies from a template. Payback under one month, build effort under a week, fragility low. Do three of those before the ambitious project and you will have both credibility and budget.
What usually loses
Anything that requires a rarely used system to expose data it was never designed to expose, and anything where the "process" is actually three people's different habits. Fix the process first, automate second.
Tools
For the winners, the tooling is boring on purpose: scheduled scripts, n8n or Make flows, and an LLM step only where text has to be understood or written. We keep a log of every run and an alert when one fails, because an automation nobody watches is a liability, not an asset.
If you want this done for your team, our consulting service starts with exactly this exercise.
