A brief lands on Monday morning and the deadline is Thursday. Somewhere in the first hour, a designer picks a direction, and everything that follows is a refinement of that early guess. Most projects work exactly this way, not because anyone thinks it is ideal, but because exploring properly used to be expensive. That calculation has changed, and the interesting consequence is not speed. It is that presenting a single route now looks like a choice rather than a constraint.
Early commitment happens for practical reasons. Comping a second visual direction meant sourcing different imagery, rebuilding a layout, finding new type pairings, and doing it all again for a third. Two extra days of work for options a client might dismiss in ninety seconds is a hard sell to anyone tracking billable hours.
The cost of that habit is hidden but real. When only one direction exists, feedback turns into a critique of your judgment rather than a conversation about fit. The client cannot say “this one, not that one,” so they say “I’m not sure about the blue,” and three rounds of colour revisions stand in for a strategic decision nobody actually made. Presenting alternatives changes the shape of the conversation, because now the client is choosing rather than judging, and choosing is a much more comfortable position for someone who does not speak design.
The change is in the rough stage, not the final one. A layout still has to be built properly, type still has to be set with care, and production files still take as long as they ever did. What collapsed is the cost of the sketch: the version that exists only to answer whether an idea is worth pursuing.
There is a discipline in keeping those sketches rough. A polished mockup invites line-by-line feedback about details that do not exist yet, while an obviously provisional one keeps the discussion where it belongs, on the direction. Some studios deliberately present exploration work in greyscale or at low fidelity for exactly this reason, and the habit is worth keeping even when the tools make it easy to produce something that looks finished.
Used at this stage, an AI picture generator is closer to a mood board than a production tool. Three routes for a campaign might mean one warm and human, one graphic and flat, one editorial and photographic, and having a rough visual for each within an afternoon means the studio can walk into the meeting with real alternatives rather than a description of alternatives. Words like “we could go more editorial” mean six different things to six people in a room. An image, even a crude one, means one thing.
This maps onto a process model most designers already know. The Design Council’s Double Diamond, developed in 2003 and now used well beyond the UK, splits work into Discover, Define, Develop and Deliver, with the explicit instruction to widen before narrowing at two separate points. The framework has always been sound and rarely followed properly, because widening was the expensive part. Anything that lowers the cost of the divergent half makes the model practical rather than aspirational.
Generating more options is now trivial, which makes restraint the actual skill. Research on choice has been clear on this for a long time. In a set of experiments published in the Journal of Personality and Social Psychology in 2000, Sheena Iyengar and Mark Lepper found that people were more likely to buy jam or chocolate, and more likely to take on an optional assignment, when offered six options rather than 24 or 30. They also reported being more satisfied with what they picked. Handing a client fifteen directions overwhelms people in a way that reads as thoroughness and behaves as delay.
There is a practical version of this for studios too. Fifteen options take fifteen explanations, and the meeting runs out of time before anyone commits to anything, so the work goes back into the queue for another week. Three works because it forces a real spread. Two options tend to collapse into a preferred one and a straw man. Four or more start overlapping, and the differences between them become too subtle to name out loud. Give each route a short name, a one-line rationale, and a note on what it would cost to produce, and the meeting becomes a decision instead of a survey.
A generated direction is a proposition, not a design. It has no type system, no grid, no defined behaviour at other sizes, and often no coherent logic behind its colour relationships. Turning the chosen route into something usable means layout, hierarchy, contrast, legibility, and a plan for how the whole thing holds together on a small screen. None of that has moved, which is why it still pays to revisit the fundamentals before production starts rather than after.
That is the honest division of labour. The tool is good at producing a plausible surface quickly. It is not good at knowing which surface serves the audience, whether the brand can defend it in twelve months, or how it degrades when someone loads it on a five-year-old phone. Those judgments are the reason the client hired a designer, and they are also the reason the exploration stage matters, since a direction chosen carelessly is expensive to abandon three weeks later.
Clients rarely remember which comp was generated in twenty minutes and which took two days. What they remember is whether the person presenting could explain why this route and not the other two. Exploring three directions is not about hedging or about producing more work. It is about arriving at the recommendation with evidence behind it, having seen the alternatives and rejected them for reasons you can say out loud.
Keep the sketches rough, keep the set small, and keep the reasoning attached to each option. The tools have made the first part of that easier than it has ever been. The rest is the same job it always was.
