One central question I ask myself is: How do we unlock value from generative AI in ideation? With the rise of LLMs, there is a huge opportunity in innovation management and product development to generate new ideas automatically and at scale.

A simplistic view is that we already derive value from LLMs just by generating ideas. An LLM can easily produce, say, 10 ideas centered around, e.g. sales tools. My issue is that the process is a blackbox. It lacks the transparancy of how each of the outputs was generated, thereby making it very difficult to steer further processes.

What would success look like in using LLM for ideation? I would argue that we derive value from LLMs when we can optimize three areas

1. Output: The generative process creates a large volume of high-quality ideas that also span a wide variety within the chosen focus area. This would directly improve the current state, which often shows uneven quality, lots of overlap, and a constrained solution space because people tend to think in similar ways.
2. Process: A structured and transparent process that records the decisions, reasoning, and steps used to generate each idea. Ideally, the transparent process and output offers an interface to change parts and variables of the output to steer the direction of the process. This would highly improve the current black-box process where outputs simply appear.
3. Organization: The workflow helps to organize and manage the inherently messy idea space. The amount and variety of the ideas makes it often impossible to have a good overview of the space. This would reveal blind spots, already covered areas and can help steer further ideation. Valuable information should easily be added to increase information density for example for evaluation processes.

I will go further into details about how we can build a Generative Ideation or a A.I-augmented Ideation with additional posts.