What to Automate First in a Small Team

Automate the mistake you already make on Tuesdays. Start with work that is frequent, boring, and easy to get wrong. Leave judgment in human hands until the rule is actually a rule.

I was the small team on the mystic-bytes cover pipeline. Manual normalize-before-build meant missed nights and bad hashes. The ROI was obvious: hours saved per month versus hours to write and keep a script. Deploy, lint, test, dependency bumps behind a CI gate, backup verification, certificate expiry: that list is toil. Ambiguous product calls and one-off migrations are not. Automating those just hides the decision in a job that fails at 2am with nobody who remembers why it exists. They reach for a platform before they have a script that already works twice.¹

Start with the work that wakes people: restore scripts, post-deploy smokes, a rollback you can press. Automation should fail loudly. Silent success that skipped a check is worse than a noisy script you still run by hand.

If you cannot describe the correct output without a meeting, do not put it in cron. Automate the Tuesday mistake first.

— JV · Dark Heart Labs.

Betsy Beyer et al., Site Reliability Engineering (O’Reilly, 2016), ch. 5 — toil is repetitive, automatable, and interrupts real work; that is the queue to drain first. ↩