We use AI at Di Marca every day. It has made parts of our work faster and some of it better, and I have no interest in pretending otherwise.
What I have become uneasy about is how much brand work now arrives looking agreed when nobody has agreed to anything.
Two versions of that have turned up in the last few months. One came in a brief. The other I sat through in a meeting.
The brief
More than once now, we have been sent a brief that was better written than most briefs we receive. Thorough, well-ordered, specific about audience and objectives, clear about what was wanted and why. Written with AI, and you could tell. There was nothing vague in it.
So we followed the brief and scoped the project.
Only later, after a fair bit of work had gone into it, did we find out that much of what was in that document had never been decided. Those weren’t calls the client had made. They were the AI’s suggestions, and the client had not read them closely enough to notice the difference.
They had not tried to mislead us. They had read the brief back, and it looked right to them, so they sent it. What they could not see, because they had not built it themselves, were the claims, requests, and assumptions the document was making on their behalf.
Why did I fall for it?
A long brief has never been proof of much on its own. Plenty arrive thick with detail and full of holes you can see from the first page.
What was different here was the pitch that came with it. Here is exactly what we want. We have done lots of the work up front so your job is easier, and by implication the budget goes further.
A document of that size used to cost something to make. It meant somebody had gone away and thought hard, had the debate internally, and come back with positions. The effort was the proof, so you could take the claim at face value and be fairly safe in doing so.
I wrote a piece a while back about costly signalling, the idea that we trust signals precisely because they are expensive to fake. A billboard on the M1 tells you something because you cannot get one for nothing.
A document like that costs almost nothing to make now. The signal is still being sent, and it no longer carries the information, or the value, that it used to.
The client cannot spot it either. They are reading the same document I am, and it looks just as settled to them. Most people do not write enough briefs to know what a good one leaves out. So it is on me to catch it.
It is not one bad brief
BetterUp Labs and Stanford’s Social Media Lab surveyed 1,004 full-time desk workers about the AI-generated work being passed around inside their organisations. 40% had received something in the previous month that looked like finished work but wasn’t. The researchers gave it a name in Harvard Business Review: Workslop. Content that appears polished but lacks the substance to meaningfully advance the task.
Each instance took an average of 1 hour and 51 minutes to sort out, roughly 20 minutes longer than it would have taken the sender to do the work properly in the first place. Across a 10,000-person business, that runs to more than $9 million a year.
The finding that stuck out for me is about trust. Around half the people who received workslop rated the sender as less capable, less creative and less reliable. 42% trusted them less.
Brand work runs on that trust, because a strategy only does anything if the rest of the business believes it enough to act on it.
The meeting
I sat in a meeting recently where a business presented brand work they had built themselves, with AI doing most of the work. Positioning, messaging, audience thinking, the lot. There was a lot of it and it was not lazy. Someone had put in real hours.
Then the questions started.
The people who had done the work could not explain what they had arrived at. They had produced the material without ever having the debate that normally comes with producing it.
The others in the room had not been involved at all. They asked two questions.
“I wasn’t aware we had agreed to cut this language.”
“Are we all aligned on how much change this means for the future?”
Neither question is about the work. Both are questions about when the decision was made, and by whom.
The person who developed the deck struggled to take them through the reasoning, because there was no reasoning to fall back on. There was an output.
The thinking never happened
When a business works out its positioning properly, which is to say who it is for, what it stands for, and why anyone should choose it, two things come out at once. There is the document, and there is a group of people who understand why the document says what it says.
The second is what the debate produces. Spend a few hours in a room working out who the business is for, and everyone leaves with a view on it.
AI hands you the document on its own. You get the conclusion with no path attached to it, which is why a client cannot always tell their own decisions from their own wishes, and why a room full of people can be shown a finished strategy and have no idea when they agreed to it.
Most of the writing on this problem treats it as an individual one. The person who leaned on the tool ends up not understanding their own document, and some poor colleague downstream has to fix it. True enough, and for brand work it is half the story. A positioning nobody can explain does not get fixed downstream. It just gets quietly ignored by everyone who was not involved, which is most of the business. Outputs without reasoning are only going to widen the strategy execution gap that so many businesses already have.
Get the right people in the room
We make sure the thinking happens by putting everyone in a room. Face to face, our people and the client’s, working through the decisions together over a day or two.
We call it a workshop, which does it no favours. It sounds like a day of activities that ends with a photo of a wall of Post-its. Ours is a day of decisions.
It is the core of how we do strategy, and I expected AI to weaken the case for it. It has done the opposite, which I did not see coming.
This is where I would normally be accused of defending my own patch, so rather than just telling you workshops are good, here is what actually happens in one.
Clients want the document. It is the thing you can see and hold, so it is fair enough to assume that producing it is the work.
They were never paying for the document, though. They were paying for the collaboration and process that enabled it, and for the alignment that only comes out of hashing it out together.
Our workshops are slow on purpose. You put the right people in a room; you make decisions in an order that builds, because each depends on the one before, and half of what comes up gets discarded in front of everyone.
What you let go of can be more important than what you leave in. When a room watches an option get killed and hears why together, the option that survives carries the reason with it. A document contains only the survivors. It tells you nothing about what was considered and rejected, which is most of what you need to defend it later.
What you have at the end is a decision, and a group of people who can explain how they got to it. That is the part that transfers. Someone who was not there can be walked through the reasoning by someone who was.
Somebody else has to be able to say it
This matters especially for brand, because a brand decision is worthless if only the people who made it understand it. The sales team has to carry it, whoever answers the phone has to carry it, and you have to describe the business to someone at a barbecue. It is what people say about you when you are not in the room, which means somebody other than you has to be able to say it.
Which makes the real decision who is in the room. The people who will have to live with the answer, the ones who will quietly veto it later if they were never consulted, and the one person who thinks you have got it wrong. Before you book anything, work out who needs to explain this in six months. Those are the people who need to be there when it is decided.
Where we use AI
Preparation, mostly, and it is very good at it. Sorting interview transcripts. First-pass audits of a category. Finding patterns across research we would otherwise read twice. Twenty versions of an idea so we have something to react to rather than a blank page.
It has shortened all of that considerably. It has not shortened the deciding by a single minute, and I do not expect it to, because the deciding was never a processing problem.
We have changed one thing on our side too. When a brief arrives now, we go through it and ask which decisions you have made and which you want our help making. It is an awkward question to ask a client who has just handed you a polished document. It saves a great deal of trouble later.
What has not got faster
What it comes down to is what happens to a decision when nobody has actually had to make it. A document can now be produced without anyone in the business having taken a position, and it will read exactly like a document produced by people who did. That is a new problem, and the person who pays for it is whoever is left holding the brand.
Which means the things worth protecting are the ones that have become easy to skip. Getting the right people in the same room. Making decisions in an order that builds. Discarding options out loud, so the ones that survive carry their reasons with them. Making sure the people who have to live with the answer were there when it was reached.
None of that has got any faster, and I don’t think it will.
Anyone can have the document in an afternoon now. Agreeing on what it means, and what it will change, takes as long as it ever did.
