I use AI every working day. Claude, ChatGPT and Google Gemini sit open alongside everything else I do across five brands. I’m not going to perform anxiety about that, and I’m not going to pretend the tools run the marketing. They’re very good at some jobs and still poor at the four that matter most.
It can’t read the room
AI learns from everything, so by default it writes for everyone. That’s exactly the problem when your customer is someone specific. Over my career I’ve marketed excavators to site foremen, safety boots to tradespeople walking into a store, a law firm’s expertise to its clients, and motorcycles to Harley-Davidson riders. Each group has its own vocabulary and a finely tuned detector for copy written by someone who doesn’t know them.
A model can write a grammatically perfect ad for a machine. It won’t know that a foreman rolls his eyes at “unleash your productivity”, or that trust in that market is built on specifications, uptime and the local service team rather than metaphors. Knowing when to strip the polish off and speak plainly comes from time spent in the yard, the store and the showroom. AI can match the syntax of a market. It can’t feel its texture.
It can’t make the call
AI is excellent at generating options. It doesn’t choose between them, and it doesn’t live with the consequences. When a campaign’s cost per inquiry jumps mid-flight, somebody has to decide: cut the budget, change the creative, or hold because the sales cycle is longer than the reporting window. A tool can lay out the trade-offs in seconds. It doesn’t answer to the owner or own the budget.
At AdvanceQuip, two of the excavator models we were advertising sold out partway through a campaign. Pulling their ads and extending the finance offer across the rest of the CASE range wasn’t a hard analytical problem. It was a judgment about stock, sales capacity and what the business needed that month, made by people who understood all three.
It can’t earn trust in the room
Marketing plans don’t execute themselves. They need people to agree to them. Moving budget between brands, changing a long-running program or asking the sales team to help produce content all require someone to make the case and answer the questions nobody prepared for.
When I reported to the partnership board at AWS Legal, the partners weren’t approving a document. They were deciding whether to trust the person presenting it. AI can draft the outline of a proposal. It can’t build the relationships that get the proposal approved and delivered.
It doesn’t know when to stop
Ask a model for a campaign brief and you’ll get six headings and three pages. It struggles with economy. Often the most useful brief is the shortest one, and the right launch is a three-line email rather than an integrated campaign. Knowing when something is good enough to ship is a taste filter built from years of watching what customers actually respond to.
The tools have removed a lot of friction from the mechanical parts of the job. The core of it hasn’t moved: who we’re for, what the brand stands for and what we fund next. If anything, that judgment matters more now, because everyone has access to the same average answer.
Read the Marketing, Reforged sample for a longer look at how marketing decisions get made and kept.