An algorithm that thinks like a creative director

If you ask an LLM to brainstorm business ideas, you’ll get very familiar, formulaic answers. Not bad ideas, necessarily – but not original. 

And that’s not surprising. Generative AI models are built to predict plausible responses based on patterns in data, which means they tend to gravitate toward concepts that already resemble things in the world.

If you’re hoping to use AI as a catalyst for innovation, this means you hit its limits very quickly. But new possibilities for AI-powered idea generation are emerging, bit by bit. 

Engineering creativity 

A recent research project set out to explore whether machines could be designed to think about creativity in a more deliberate way. Instead of relying purely on statistical pattern-matching, the researchers developed a computational model intended to mimic certain elements of human creative reasoning. Their goal was to generate real innovation opportunities – the kind of ideas that might lead to new products or services, or help develop a new business model. 

The system was built around five internal functions designed to increase novelty while maintaining practical usefulness. 

In essence, the researchers attempted to encode a set of creative strategies into software, giving the system mechanisms to explore unusual combinations of ideas and move beyond the most obvious solutions. This caught our attention – because it’s the kind of function the founders and business leaders we meet at LEAP could really benefit from.

Testing the model in the real world

To test the approach, the model was applied to a real innovation challenge in the hospitality sector. Its task was to generate potential opportunity ideas for the industry, which were then evaluated alongside ideas produced by widely used AI tools including ChatGPT-4o and Google’s NotebookLM.

According to the study, the outputs generated by the computational creativity model were judged to be more novel (and in some cases more useful) than those produced by commercial systems. 

The results are only an early step, and the researchers do note that not every part of the model contributed equally to its performance. Still, the experiment suggests that creativity might not be an inexplicable spark of inspiration. It may also be a process that can be partially engineered. 

The rise of computational creativity

The research sits within a growing field known as computational creativity – which aims to build computer systems capable of simulating aspects of human creative thought. 

For decades, researchers have experimented with algorithms that compose music, write poetry or generate visual art. The emergence of LLMs has accelerated these efforts. 

A related idea already appears in engineering, through generative design – where software explores thousands of possible solutions to a design problem before humans choose the most promising one. Instead of replacing designers, the system expands the landscape of possibilities and speeds up the problem-solving process. 

And this new research applies this principle to innovation itself.

Building creative partners for the future 

Today’s AI tools largely act as assistants. But systems built around creativity frameworks could behave differently. 

Instead of just responding to prompts, they might actively search for unusual combinations of ideas, and surface possibilities that human teams might overlook.

If that happens, the role of AI in creative industries will evolve again. Generative models showed that machines could produce content at scale; computational creativity hints that they might also help design the thinking behind that content.

So the most effective creative teams of the future might look more like hybrid studios. Humans bring judgement, taste and direction. Algorithms roam widely through the landscape of ideas, returning with unexpected combinations worth exploring.

When that happens, the machine becomes a tireless brainstorming partner – one that never quite runs out of possibilities to test.

Join us at LEAP from 31 August – 3 September 2026 to understand how creativity is starting to be engineered into AI systems.