The most expensive automation mistake isn't a bug. It's automating the wrong thing. Polishing a process that runs twice a month feels productive but returns almost nothing. Before we write a line of code, we run a short audit to find the work that actually deserves it.
Map the work as it really happens
Not the documented process, the real one. We sit with the people doing the work and trace it step by step: the tools they switch between, the copy-paste, the "and then I just email Sarah" handoffs. The hidden manual steps are usually where the time goes.
Score each candidate
We rate every repetitive task on four axes:
- Frequency: how often does it happen?
- Time: how long does each instance take?
- Error rate: how often does it go wrong, and what does that cost?
- Feasibility: how cleanly can it be automated with today's tools?
Frequency times time gives you raw hours. Factoring in error cost and feasibility tells you where to start.
Start with the compounding win
The best first project is high-frequency, time-consuming, error-prone and technically tractable. It returns hours immediately, builds trust in automation, and creates a foundation the next project can reuse.
Be honest about what not to automate
Some work needs human judgement, happens too rarely to justify the build, or would cost more to automate than it returns. Saying so is part of the job. Good automation is targeted, not total.
The output of the audit is a ranked list and a clear first move, so the first project pays for itself and earns the right to a second.