AI for marketing teams: when everyone has the same machine
The function that adopted these tools fastest is the one now paying for it, because a house style built out of the default voice is the least distinctive thing a brand can own.
Every one of your competitors now writes their first draft on the same machine you do, and the drafts have started to rhyme. AI for marketing teams took hold faster than in any other function, for the plain reason that the work is made of words and the tool makes words. It also produced the problem that now defines the function.
That costs money, not pride. Sameness is the one thing a brand cannot charge for. Whatever edge existed in producing more copy more quickly has already been handed to your competitors, free of charge.
Write the voice brief once
The most useful hour I have spent with a marketing team was the one where nobody wrote anything. We put their standing brief up on the screen, all eleven words of it, and I asked who the piece was for. Four people answered and no two answers matched. One of the answers was a job title.
Anthropic's published prompting guidance leads with clarity, examples and role, and in this function those three do nearly all the work. The mechanics of assembling a brief are set out in the piece on writing a prompt that works.
A voice brief the model can follow has to name things. It says who is speaking and in what role, and who is being spoken to, described as a person rather than a segment. It states what the piece must not do, as prohibitions your team would actually recognise. Then it pastes in two or three pieces of your own published writing to match, which achieves more than any adjective ever will. Confident but approachable means nothing to a model, and if you are honest about it, not much to a new starter either.
Once that brief exists in writing, repurposing stops being a risk. One well-argued point of view can become the conference talk, the long post, the customer email and the sales one-pager without losing its shape, because the brief keeps the voice steady across all of them. The mistake is repurposing before the thinking is finished, at which point you are distributing a shrug across four channels.
Readers can hear it now
Efficiency is the weakest reason to take any of this seriously. The real one is that a particular cadence of writing has become recognisable to ordinary readers, and recognisably cheap. The tells are consistent: neat triads, a paragraph that opens by dismissing an argument nobody made, a closing line that restates the opening in a slightly grander register. Two years ago that read as competent. It now reads as unattended.
That shift is a liability, and it falls heaviest on the brands that moved first and fastest. What is left to compete on is whether you have a view somebody would notice was missing.
Drafting was never the hard part
Where teams come unstuck is treating the output as ninety per cent finished when it is closer to a structural sketch with the interesting parts left out. The distance between that sketch and something you would put your logo on is where a marketing team either wins its time back or loses its voice by degrees. Your people team is further down the road on written rules than you are, and what they are doing with the same technology is a fair template when you come to write your own.
Read the next thing you approve out loud, the whole of it. If nobody in the room can say whose voice that is, the machine did not cause that, it only made it cheaper to keep producing.