The Automation Audit: Finding What's Actually Worth Automating
Not everything should be automated. A framework for identifying high-ROI automation targets before writing a single line of code.
The most expensive automation mistake is automating the wrong thing. Building a sophisticated pipeline for a task that takes 10 minutes per week — and that you'd likely stop doing anyway — is waste. Before any technical scoping, I spend time on a structured audit.
The three axes I use: frequency (how often does this happen), consistency (does the process follow the same steps each time), and cost (what's the actual time or error cost of doing it manually). High-frequency, high-consistency, high-cost tasks are the obvious targets. The interesting ones are low-frequency but high-consistency — occasional processes that, precisely because they're infrequent, get done inconsistently and slowly.
The output I'm looking for isn't a list of 'things to automate.' It's a ranked list with a rough ROI estimate and a confidence band. This gives the client a clear picture of where to invest: quick wins first, complex projects with high certainty second, speculative projects last.
One thing I've learned to watch for: processes that look consistent but have hidden variance. 'We process invoices the same way every time' often means 'same way, except when the vendor format is weird, or it's end of month, or Sarah is on leave.' Document the exceptions before you automate — they're the part that will break your system.