People ask me whether AI has changed my job. The honest answer is that it changed my hours, not my judgment. I still decide the strategy, the offer, and the angle. What I no longer do is spend an afternoon on the mechanical work around those decisions.
Here is what a normal week actually looks like, tools and all, without the hype.
Two models, two jobs
I use both ChatGPT and Claude, and I do not treat them as interchangeable.
Claude is where I do thinking work: long briefs, structured analysis, reworking a messy strategy doc into something a client can read. It holds a long context well and tends to stay on the rails when I give it a lot of material.
ChatGPT is where I do fast, high-volume production: forty ad variations, a batch of subject lines, quick reformatting. It is quick and it plays nicely with images and spreadsheets.
That split is personal, not gospel. The point is that I picked a default for each kind of task instead of opening whichever tab was already there.
Mornings: triage and drafts
My first AI task most days is the inbox and the calendar, in plain words. I paste the week's numbers and ask for a short summary of what changed and what needs a decision. It is not magic. It just removes the friction between "I have data" and "I know what to do today."
Then come the drafts. If a client needs a landing page outline, a newsletter, or three ad angles, I write the brief myself (audience, offer, tone, one thing to avoid) and let the model produce the first version. The brief is where my expertise lives. The draft is just clay.
A good brief gets me a draft I can edit. A lazy brief gets me beige copy I have to rewrite from scratch, which is slower than not using AI at all.
Catalogs: the unglamorous win
Product catalog work is where AI quietly saves me the most time, and nobody writes about it because it is boring.
When a store sends me a spreadsheet of 300 products with thin or duplicated descriptions, I used to budget days. Now I set up a clear template (benefit-led, keyword-aware, fixed length, no superlatives) and generate first drafts in batches. I read every one, fix the handful that miss the brand voice, and the catalog is ready for a human eye instead of a blank page.
The same goes for translations. I sell into a few markets, and machine translation now gets close enough that my job is editing for nuance rather than translating from zero. Close enough, with a human pass, beats perfect and never finished.
Campaigns: variations, not ideas
This is the line I hold. I do not ask AI for the big idea. The hook, the positioning, the reason someone should care, that is the part I am paid for and the part models are still mediocre at.
What I do ask for is variation around an idea I already have. Give me ten ways to phrase this hook. Rewrite this for a colder audience. Tighten this to fit the character limit. That turns one good concept into a real testing plan, which is what actually drives performance on Meta and Google.
The mental model that works for me: AI is a brilliant copy assistant and a forgettable strategist. Use it accordingly.
Where I refuse to use it
A few places, on purpose.
- Final numbers and claims. Anything a customer could hold me to gets checked by a person, every time.
- Sensitive client data. I do not paste things into a tool that I would not be comfortable seeing leak.
- The actual decision. Whether to kill a campaign, raise a price, or pivot an offer is mine. The model can lay out the trade-offs. It does not get the call.
The real productivity story
The framing I dislike most is "AI replaces marketers." In practice it does something less dramatic and more useful. It deletes the busywork that used to sit between me and the parts of the job I am good at.
A catalog that took three days takes one. A testing plan that I used to skip because I was tired now gets built. A report that ate my Friday afternoon writes its own first draft.
None of that is a robot doing my job. It is me doing my job with the boring 60 percent handed off, so the important 40 percent gets my full attention. That is the whole pitch, and after a couple of years of living inside these tools, it is the only claim about them I fully trust.