Two years of generative tools in the studio has changed our process considerably and our briefs almost not at all. That gap is worth explaining, because a lot of clients now arrive with an assumption that AI should make the brief lighter, faster, or unnecessary.
The opposite is true. When execution gets cheap, the brief becomes the most valuable document in the project, because it is the only remaining thing that determines whether the output is right rather than merely plausible.
What actually changed
Three things, concretely.
Exploration got wider. We can now put twenty visual directions in front of a client in the time it used to take to produce four. Not finished work – directional, imperfect, enough to have a conversation about. The value is not the images. It is that the client can react to breadth instead of imagining it.
The middle of the process compressed. Rough boards, placeholder imagery, first-pass copy variations, storyboard frames for a motion piece. Work that used to consume days now takes hours. That time has not disappeared from projects. It moved to the front, into thinking, and to the back, into craft.
Bad work got faster. This is the part nobody advertises. Generative tools produce competent, generic output very quickly. A studio without a strong point of view now produces average work faster than before, and charges the same for it.
What did not change
The strategy. No tool knows your market position, your margin structure, why your last launch underperformed, or which competitor your best clients actually compare you to. It will generate confident-sounding positioning statements based on nothing. Positioning still comes from evidence, argument and judgment. We wrote about what that work involves in what brand strategy actually is.
The decision. Twenty directions is not better than four unless someone can say why one is right. Generative tools multiply options and do nothing to help you choose. If anything they make choosing harder, because the volume creates an illusion of thoroughness.
Craft at the final ten percent. The difference between an image that is nearly right and one that is right is still hand work. Kerning, weight, timing, the specific way a mark sits at 16 pixels. This is where projects are won and it is exactly where the tools are weakest.
Ownership and provenance. More on this below, because it belongs in your brief now.
What this means for the brief you write
If execution is cheap and judgment is expensive, a good brief front-loads judgment. Practically, that means five things.
1. Be much clearer about the problem, and looser about the solution
The old temptation in briefs was to over-specify the output – “we need a new logo, three social templates and a video”. That was never great, and it is now actively wasteful. Specifying deliverables is the cheapest part of the project to change.
State the business problem instead. What is not working, for whom, and what would be different if it were fixed. Give the studio room to argue that the answer is not the thing you named. Half the value of a strategic partner is the recommendation you did not ask for.
2. Bring evidence, not references
Reference imagery has always been a mixed blessing. It is now close to a liability, because a folder of references is precisely the input a generative tool consumes best, and the fastest route to work that resembles other work.
More useful in a brief:
- What your best customers actually say about you, in their words.
- What you lose deals on.
- Which competitor you are most often confused with, and why that hurts.
- What the business must be able to say truthfully in two years that it cannot say now.
None of that is generatable. All of it changes the output.
3. Say where you stand on AI in the work
This is new and it belongs in every brief now. Studios use these tools. The question is where, and whether you are comfortable with it.
Reasonable positions we have seen from clients, all defensible:
- No generated assets in final deliverables, but tools are fine for internal exploration. Common in regulated sectors and among clients with strong craft values.
- Generated imagery acceptable for concept and social, not for the identity itself. The most common middle position.
- Fine anywhere, disclose it. Increasingly common with fast-moving consumer brands.
- Photography and illustration must be original and licensed. Usually where there is a real risk of a provenance question later.
Whichever you pick, write it down. The failure mode is discovering the disagreement at delivery, which is expensive and sours an otherwise good project.
Ask about ownership at the same time. Who owns the output, what warranties the studio gives on originality, and what happens if a third party raises a claim. A studio that has thought about this will have a clear answer. One that has not is a risk you are carrying.
4. Budget the thinking, not the pixels
If you are used to costing projects by output volume – so many pages, so many assets, so many revisions – that model is drifting out of alignment with where the work is.
Expect a well-run project now to weight more of its budget toward discovery, strategy and the decision points, and less toward pure production. That is not a studio finding a way to charge more for less. It is the cost following the value.
The corollary: be more suspicious of quotes that are cheap because production is cheap. Fast, generic execution has never been more available, and it has never been worth less. This is one of the ways the cheap-website trap has got worse, which we set out in the hidden cost of a cheap website.
5. Keep the approval structure tight
Generative tools make it trivially easy for anyone in the organisation to produce an alternative and bring it to a review. We have sat in sessions where a stakeholder arrived with their own generated version of the logo.
That is not automatically bad – sometimes it clarifies what someone was struggling to describe. But it is corrosive if the approval structure is loose, because it turns a design review into a competition between a considered proposal and a fast approximation.
The remedy is the same as it always was, only more urgent: one decision-maker with final say, and a review process that evaluates work against the brief rather than against preference. Our brief template has a section for exactly this, and it earns its place more than it used to.
How we use these tools, plainly
Since we are asking clients to be explicit, we should be.
We use generative tools for early exploration – rough directions, mood, compositional options, first-pass storyboard frames for motion work. We use them to draft and stress-test copy variations, never to write final copy. We use them to speed up mechanical production tasks like format variations and rough cut-outs.
We do not ship generated imagery as a final brand asset unless a client has explicitly agreed to it. We do not use them for strategy. And we tell clients where a tool was involved, without being asked.
That policy is not a moral position. It is a quality one. The tools are excellent at producing something that looks like the answer, and that is a dangerous thing to have available when you are tired and the deadline is Thursday.
The thing worth holding on to
The studios that get worse over the next few years will be the ones that use these tools to produce more, faster, at the same price. The ones that get better will use the recovered time to think harder about fewer things.
As a client, you can tell which kind you are talking to from one question: when you describe your problem, do they start generating, or do they start asking? Everything else follows from that.
If you are about to brief a project and want a view on how to write it so it survives contact with a fast-moving toolset, send a short note describing what you are trying to change in the business. We are happy to look at a draft brief and say where it is under-specified.