What Automation Can and Can't Do
This is a member-only chapter. Log in with your Signal Over Noise membership email to continue.
Log in to readModule 1 · Section 4 of 5
What Automation Can and Can’t Do
Let me be honest about what AI automation actually fixes here.
It does not write your newsletter for you. If that’s what you’re hoping for, this course will disappoint you. Issues that are entirely AI-generated are detectable — not by any magic filter, but by the fact that they have no perspective. They summarise. They list. They don’t have a view.
What automation fixes is the overhead that doesn’t require your judgement. Research — gathering sources, extracting relevant points, building a brief — is largely mechanical. You still decide which topics matter and why. The AI does the reading.
First-draft generation is something AI does well when you give it the right inputs: a research brief, a voice profile, a clear structure. The output isn’t publishable as-is, but it gives you something to edit rather than a blank page to fill. Editing from a draft is faster than writing from nothing.
Review and quality checking — catching AI slop phrases, flagging sentences that don’t sound like you, testing whether the opening hooks — can be automated without removing your judgement, because the automation flags issues rather than silently accepting them.
Distribution — scheduling broadcasts, managing subscriber tags, tracking whether people opened — is entirely automatable. This is where Kit.com and the Kit CLI come in.
The time that stays manual: deciding what to write about, reviewing the draft the agent produces, making the editorial calls about what stays and what goes, and reading the final version before it sends. That’s thirty to forty-five minutes rather than four hours.