Great advertising starts with great ideas. Whether you’re launching a campaign, shaping a content strategy, or crafting a fresh message, it all begins with ideation. But here’s the challenge: we only recognize the path to a great idea in hindsight [1]. That means discovery must be divergent—open-ended, exploratory, and loosely constrained.
This is why marketers turn to brainstorming sessions and creative workshops. These formats are designed to expand thinking, encouraging teams to explore a wide range of directions before narrowing in. Still, their success depends heavily on who's in the room and how the process unfolds.
The problem with humans
Traditional ideation is difficult to scale and notoriously inconsistent in outcome. Even when intended to promote divergence, sessions may collapse into premature convergence due to misaligned reasons. Groupthink sets in. Strong personalities dominate. Time pressure forces conservative choices.
Assembling truly divergent thinkers is also difficult. Internal groups often share similar backgrounds and blind spots. And with the pace of change in media, culture, and consumer behavior, even well-composed teams struggle to stay fresh.
Now that everything is “generative”, it’s tempting to imagine replacing human brainstorming with AI. After all, it’s supposed to be cheaper, faster, and better, right?
The problem with AI
It’s tempting to dream of handing your campaign brief to an AI, grabbing a coffee, and returning to brilliant ideas. But that dream is misleading. Plain GenAI models lack the essential mechanics of good brainstorming: divergence, criticism, and recombination.
Tools like ChatGPT support a wide range of tasks, but they weren’t built to replicate creative ecosystems. Their capabilities reflect the structure of their training, and the kind of open-ended, collaborative brainstorming we’re exploring here isn't part of it [2]. Even if it were, marketers would still face the challenge of crafting prompts far more detailed than a typical brief, requiring them to articulate things they often only understand tacitly. And that, in itself, is no small feat.
So it seems we have no choice but to leave only the routine tasks to AI: filling in the blanks once the core ideas are set, drafting that email, polishing the copy. That’s where it shines, one might say. Freed from the tasks AI can automate, humans can work on their own to improve what AI can't.
Except we can do better than this.
A creative symbiosis
At modell.ai, we see a powerful opportunity to scale and enrich creative exploration—if humans and machines work in concert.
By modeling digital creators, influencers, thought leaders, and artists, AI can simulate ideation sessions with a crowd of divergent voices. But this modeling must go far beyond mimicry. Imitation not only fails to serve our creative goals, it also brings ethical concerns and risks infringing on intellectual property. What matters anyway isn't copying outputs, but capturing the essence of what drives and repels these creators [3]. This isn’t about parroting; it’s about replicating the generative conditions behind real human creativity.
Imagine gathering a virtual room of the most original creators: each informed by your brief, yet exploring their own distinct, authentic paths. And crucially, marketers remain in the driver’s seat: guiding and steering the process, in real time or asynchronously. You might step away for that coffee, but on your return, you'll be spotlighting promising ideas and shaping the terrain for the next ideation round.
The result is a living stream of ideas: constantly evolving, recombining, and responding to feedback. It produces unexpected, high-potential directions that would be difficult to reach otherwise. Marketers are freed from the bottleneck of generation and can focus their own creativity and energy on context refinement, decision-making, and execution.
By modeling a distributed, diverse collective of creators, we avoid collapsing into bland averages. And by empowering marketers to apply selective pressure, we build a symbiotic human+AI system that pushes outward: surfacing the novel, the authentic, and the relevant.
In a world where creative timelines are tight and attention is fleeting, this shift—from purely human ideation to augmented discovery—isn’t just helpful. It’s transformative.
[1] Why Greatness Cannot Be Planned: The Myth of the Objective by Kenneth O. Stanley and Joel Lehman.
[2] This assertion is based on our understanding of how current models are trained. Existing training algorithms do not optimize for genuinely divergent exploration under selective pressure. Web-scale pretraining may contain snippets of online brainstorming, but combined with supervised fine-tuning it mostly teaches models to mimic the surface form of ideation rather than the mechanisms that make brainstorming effective. Reinforcement-learning–based reasoning likewise falls short: its rewards target correctness, whereas creative ideation rarely has a single “right” answer.
[3] There are several ways to achieve this. One approach is to capture a creator’s style, background, and intent by analyzing their public work.
Through focused partnerships and consulting, modell.ai is helping engineer this next frontier in creative intelligence.
We are currently tackling the challenge of boosting AI creativity in partnership with the Laboratory of Computational Intelligence at ICMC/USP.
Curious about other ways to collaborate? Let's talk!