From brief to concepts without a blank prompt

Generic campaign concepts are almost always a brief problem. The model generated against a vague request — "come up with some campaign ideas for this brand" — with no audience specifics, no positioning context, no examples of what good has looked like for this client. The output was generic because the input was generic.

Loading the brief before the concepting session starts is the single highest-impact change in AI-assisted creative work.

Why AI concepts feel generic

The model isn't bad at concepting. It's bad at concepting without context — which is how most AI concepting sessions start.

A session that opens with "generate ten campaign concepts for a B2B software company" gives the model nothing specific. The brand has no positioning history in the session. The audience has no specific objections or beliefs. The campaign has no objective that distinguishes it from any other campaign in the category.

The output reflects all of that. Ten generic concepts that could apply to any B2B software company, in any industry, at any competitive moment. All plausible. None of them actually interesting.

The instinct is to add more to the prompt — "make them more creative," "try a different angle," "be bolder." This rarely works because the model still has nothing specific to be bold about.

What to load before the first concepting session

Four things that change the concepting output:

The audience in specific terms. Not "marketing managers at mid-market companies." The actual decision-maker — what they believe coming in, what objection they carry, what the campaign needs to shift in their thinking. A specific audience produces concepts aimed at a real person rather than a category archetype.

The brand's positioning commitment. What the brand is prepared to say that competitors aren't. The specific claim the campaign should advance. The territory the brand owns or is trying to own. Without this, the model generates positioning from general category content, which sounds like everyone else.

Examples of what good has looked like. Past campaigns that worked. Creative that the team considers representative. One concrete example of a concept the client approved. The model compares against examples faster and more accurately than against abstract descriptions of quality.

The campaign objective in plain terms. Not "increase brand awareness." What awareness of what — what the audience should know or believe by the end of the campaign that they don't know or believe now? A specific objective produces concepts that address it.

The concepting frame: territory, not just ideas

The most productive AI concepting sessions don't produce finished ideas — they produce territory worth arguing about.

Territory is the thematic direction a campaign could take: "the cost of the status quo" as a frame, or "what you lose without acting" as an emotional premise, or "we're the only one saying X." These are directions, not executions. They're useful because they're arguable — the team can debate whether this territory is right before committing to executions.

Brief the model to generate territory rather than finished concepts. "What frames could this campaign take, given this audience and this positioning?" produces more useful output than "give me ten campaign ideas." Territory is easier to evaluate and easier to build from.

Evaluating AI concepts: three questions that separate working ideas from filler

Not all generated territory is worth pursuing. Three questions that filter fast:

Is this specific to our brand? Could another company in this category use this concept without modification? If yes, the territory isn't ownable — the brief needs more positioning specificity, and the round needs to go again.

Does this address the audience's actual objection? Not the audience in general — this audience, with the specific belief or objection the campaign is trying to shift. Concepts that don't address the specific objection are technically on-brand but won't move the needle.

Can we execute this? Some territory is defensible in theory but can't be produced with the team's actual resources, timeline, or risk tolerance. Evaluate for executability before investing in development.

Concepts that pass all three are worth developing. Concepts that fail the first question go back to the brief; concepts that fail the second tell you the brief didn't adequately capture the audience.

Moving from concepts to execution without losing the thread

The most common place the campaign falls apart: the hand-off from concepting to execution. The team approves a direction, opens a new session for execution, and the model starts without the approved concept, the brief it came from, or the audience context that made the concept work.

The execution session should start with the same loaded brief the concepting session used, plus the specific concept direction the team approved. The model then executes from that frame rather than starting fresh.

What changes: the execution reflects the approved concept rather than being a generic execution of a general campaign type. The consistency between concepting and execution holds because the context doesn't reset.

For the campaign workflow that makes this handoff work: The AI-assisted campaign workflow, start to finish. For keeping context across a multi-week campaign: Carrying context across a multi-week campaign. For the briefing skill that makes every stage better: How to brief an AI like a senior planner.

Frequently Asked Questions

Can AI come up with campaign concepts?

Yes — against a real brief. Load the audience specifics, the brand's positioning commitment, and examples of what good has looked like for this client, and the model generates territory worth evaluating. Start without a brief and you get generic concepts that could apply to any brand in the category. The model's creative output is bounded by what it was given. Give it something specific; get something specific back.

How do I use AI for creative brainstorming?

Brief it for territory rather than finished ideas. "What frames could this campaign take, given this audience and positioning?" produces directions the team can argue about. "Give me ten campaign ideas" produces plausible but often generic options. Load the audience specifics and positioning commitment before the session starts, ask for thematic territory rather than executions, and use the output to structure a human discussion about which direction is worth pursuing.

Is AI good at big-idea concepting?

It depends on what "big idea" means. If it means a genuinely original idea the team has never seen before — AI doesn't do this. If it means generating a range of thematic territory that a human team can evaluate, argue about, and develop into something distinctive — AI does this well when the brief is loaded. The big idea judgment is human; the generation of options to judge is where AI earns its place.

What brief format works best for AI concepting?

Specific and structured over general and open. The brief should include: who this audience is in decision-context terms (what they believe, what objection they carry, what needs to shift), what the brand is prepared to say that competitors won't, what good has looked like on past campaigns for this client, and what the campaign objective is in plain terms. These four elements give the model a real frame to generate against — which is what separates useful territory from generic concepts.