AI In Design Thinking

Brainstorming With AI Prompts: Expanding Idea Volume Without Smothering Originality

A strange thing happens when you ask a machine to generate ideas: the output can feel like a firehose - yet the goal in design isn’t to drown in options. It’s to keep the mind moving while protecting the spark that makes one option worth pursuing. The paradox is that more ideas can either sharpen creativity or flatten it, depending on how the prompts shape what the AI is “allowed” to do.

This chapter explores that tension through the lens of prompt-driven AI: how carefully worded prompts can expand idea volume while keeping originality intact and momentum alive. We’ll look at where brainstorming came from, why humans tend to hit the same walls, and what AI changes in the mechanics of ideation - especially when you’re trying to design something real, not just produce novelty for novelty’s sake.

The central mystery is simple to state and harder to answer: how can a tool that makes it easy to say “more” avoid making it easy to stop thinking?

The Spark Ladder Prompt Method and the Hidden Problem of “Too Many Ideas”

To understand why prompt wording matters, it helps to know what brainstorming was trying to solve in the first place. In the mid-20th century, brainstorming popularized a rule that sounds almost quaint now: separate the generation of ideas from the evaluation of ideas. The intent was psychological. When evaluation happens too early - when you start judging an idea before it has fully taken shape - people clamp down. They become cautious. They produce fewer ideas and spend more time polishing the safe ones.

But anyone who has run a real brainstorming session knows the other problem: after a while, people repeat themselves. The room runs out of fresh angles not because creativity disappears, but because attention gets sticky. You keep circling the same concept, the same category, the same “kind” of thought. That’s where AI enters - not as a replacement for thinking, but as a way to keep the search moving when your own mental search starts to loop.

The Spark Ladder Prompt Method is built around a specific design tension: idea volume should increase, but the “direction” of those ideas should keep you climbing toward something. The method treats prompts like rungs, not like a single instruction. Instead of asking for a massive list in one go, the prompt sequence nudges the AI toward producing ideas that are meaningfully different from one another, then gradually narrows attention toward constraints that matter in your problem space.

That’s the key to the chapter’s topic: it’s not just that AI can generate lots of ideas. It’s that prompts can create a ladder structure - wider at the top, more focused as you go - so you don’t end up with a pile of near-duplicates. A prompt can also protect momentum by reducing the “blank page” pause. Humans often get stuck at the moment where they can’t tell what kind of idea would count as progress. A well-shaped prompt gives the mind a foothold, and the mind keeps moving.

The surprising part is that “more” is not automatically “better.” If an AI prompt is too open-ended, the model may respond by averaging toward common patterns - the design equivalent of saying, “Tell me what you think,” and getting back a safe summary. If a prompt is too narrow too early, you get novelty that doesn’t connect to the actual problem. The Spark Ladder Prompt Method aims to stay between those extremes: increase variety without erasing the thread that ties the ideas to one coherent design direction.

When Brainstorming Meets Prediction: Why AI Can Either Widen Search or Flatten Taste

AI doesn’t brainstorm the way a person does. A person has a memory of past experiences, a sense of taste, and an internal model of what feels plausible. AI, at least in its common chat-style form, produces text by predicting what comes next based on patterns in its training data. That matters for creativity because it means the AI is sensitive to the shape of your request.

There’s a specific failure mode that designers feel instantly: when they ask an AI for “tons of ideas,” they may receive lots of phrases that look different on the surface but share the same underlying template. In other words, the output can increase apparent volume while reducing meaningful variation. This is how originality gets flattened - not by lack of quantity, but by repetition of the same structural idea under different wording.

The scientific grounding here is simple and mechanical: language models are good at producing plausible continuations. Plausibility is not the same as originality. When a prompt doesn’t provide constraints, the model leans on what it expects a “reasonable answer” to sound like. That tends to reproduce what already looks familiar....