What Still Can't Be Automated (I Keep a List)

I keep a list. Every time something in my work gets automated, I move it off the list, and every time I hit a wall a tool cannot climb, I add to it. It is the most honest way I have found to think about AI in marketing, because it forces me to be specific instead of either panicking or pretending nothing is happening.

Here is the current version. I update it when reality changes it, not when a headline does.

Still on the list (mid-2026)

• Deciding which goal actually matters. A model will happily optimise the metric you give it. Choosing the right metric, for this business, this quarter, this owner’s actual situation, is still a human call, and it is most of the job.
• Knowing when the data is lying. Tools present numbers with total confidence. Sensing that a spike is a tracking bug, or that a “win” is seasonal, or that the sample is too small to mean anything, comes from having been fooled before. That instinct does not come in an API.
• A real relationship. The email that gets answered, the partner who takes your call, the person who links to you because they trust you and not because you asked. Automation can scale outreach; it cannot manufacture the trust that makes outreach work.
• Taste. Knowing that a page is technically fine and still bad. Knowing which sentence to cut. This is the slowest thing to fall, and I suspect it falls last.
• Physical-world truth. Whether the shop is actually open, whether the product is actually good, whether the claim is actually true. Someone still has to go and check.

Moved off the list (things that used to be here)

• Drafting a first version of almost anything. Gone. A model does this faster than I do, and I now start most writing by arguing with a draft instead of a blank page.
• Summarising long documents and pulling structured data out of mess. Gone, and I do not miss it.

Why I keep it

Because “AI is coming for your job” is a useless sentence and “AI can’t do what I do” is a lie. The truth is a moving line, and the only way to stay useful is to know exactly where it is this year. So I write it down, and I let it embarrass me next year when part of it has moved.