ATOMENOSThe wrong first question
Most automation projects begin with a tool already chosen. Someone has seen a demo, read a case study, or heard a competitor mention "AI agents," and the project starts from there: how do we fit this tool into what we do?
That question guarantees a mediocre outcome. It optimizes for adoption of the tool, not removal of the friction. Teams end up automating the wrong step, or automating a step that shouldn’t exist at all.
What actually predicts success
In every automation project that has held up six months later, the first two weeks looked the same: someone sat with the team doing the work, watched them do it, and asked why each step existed.
Half the time, the honest answer was "because that’s how we’ve always done it." Those steps are the ones worth removing entirely, not automating faithfully.
A smaller, better starting point
The projects that succeed start narrow: one workflow, one team, one measurable outcome. Not a platform. Not an "automation strategy." One thing, done well enough that the team asks for the next one.
Automating the wrong step faithfully is worse than automating nothing at all.