Running a store alone means every job is mine. Planning, content, metrics, deadlines — the juggling isn't a distraction from the work; some days it is the work. So today I built myself some help: two small assistants I configured from scratch, one for marketing, one for project management.
Setting them up turned out to be closer to hiring than to engineering. I wrote each one a job description — really wrote it, the way you would for a person. The marketing assistant's whole job is ideas: content angles, ad copy, what's moving right now. The other one is a project manager, and its job is to be obsessive about timelines so I don't have to hold every deadline in my head.
The craft is in the instructions. "Give me a campaign" gets you filler. "A campaign in a playful, conversational voice, for people who love their dogs more than most people" gets you a starting point. Then you test, adjust, and repeat. They're eager interns — genuinely useful once guided, and in need of regular reminders about the bigger picture.
But the more instructive part of the day was watching automation fail — other people's, aimed at me.
I tried to boost a post on Instagram with a trending audio clip. Automated moderation flagged the audio. Fair enough; I picked another track. The system never registered the change — it just looped me back to the same rejection, around and around. Somewhere, a workflow was designed without ever imagining a real person who might need to change a selection.
Then Google's automated review labeled my site "potentially misleading" — no specifics, no path to a human, just a verdict and a closed door. After twenty years in UX, this is the thing I can't unsee: good automation isn't about the happy path. It's about anticipating confusion, frustration, and the edge cases a script never meets.
So both things are true at once. My assistants saved me hours today, and I'll keep them. And every system I touched that ran without a human watching it eventually trapped someone — today, me. The difference isn't the automation. It's whether anyone is standing behind it, willing to hesitate. Automation never doubts. Doubt is the part I'm here for.
Comments