Automate the Boring 80%: Finding Your First AI Automation Win
Companies usually approach AI automation the way people approach gym memberships in January: enormous ambition, no plan, quiet abandonment by March. The failure pattern is choosing a first project by impressiveness instead of fit. The right first automation is almost always boring, high-volume, and already annoying everyone who touches it.
The fit test
- Volume: it happens dozens of times a week or more — enough to measure and worth the setup.
- Structure: the inputs arrive in recognizable shapes (invoices, forms, order emails), even if messy.
- Rules: a competent new hire could learn the decision in a week — which means a model can too.
- A number: hours spent, error rate, or backlog size exists today, so the win will be undeniable.
The usual suspects
In practice the first win lives in one of a few places: invoice and PO processing in finance, inbox triage in customer service, data re-keying between systems that don't talk, or document intake in operations. These are unglamorous precisely because everyone has adapted to them — which is why 1,000 recovered hours can hide inside a process nobody thought to question.
Why boring wins compound
A first project that ships in six weeks and provably returns hours does something a moonshot can't: it converts the skeptics. Finance sees the number. Operations trusts the exception handling. The second and third automations get approved in a hallway conversation instead of a steering committee. Momentum, not magic, is how companies end up genuinely automated — and momentum starts with a boring, measurable, honest first win.