Real human data to generate and predict real innovation success
We are all uniquely human. As consumers, our decisions are complex, emotional, contextual, and often irrational. Although artificial intelligence (AI) makes new product development faster and easier than ever before, off-the-shelf generic models can distort or misrepresent consumers’ realities.
In this paper, we discuss the practice of training AI models with real consumer data, to capture the essence of what drives real consumer behavior, and to generate and predict better innovations. Without connecting with real humans, even the most powerful algorithms will not be sufficient to guarantee innovation success.
The headline of a recent Newsweek article, “Artificial Intelligence: the 21st century Gold Rush”, perfectly captures today’s sentiment: large language models (LLMs) like ChatGPT have turbocharged interest in AI and inspired thousands of companies to dive headfirst into this space.
While nearly every industry has or is beginning to adopt AI, optimal applications for new product development are particularly distinct. The following pages delve into better applications of generative and analytical AI to power innovation success.
In a typical innovation process, an ideation phase is followed by an evaluation phase. AI can be leveraged in both phases:
- In the ideation phase, the divergent capabilities of generative AI can be leveraged to develop new product ideas.
- In the evaluation phase, the convergent capabilities of analytical AI can be used to predict their market potential.
AI presents an opportunity to improve the speed, and potentially, the success rate of new innovations, and how we go about doing this will determine whether we succeed. In both applications, the data used to train AI is critical.
New product ideas are more likely to succeed if these two phases are grounded in data reflecting consumers’ intrinsically human needs and desires. This data needs to be timeless, or at minimum, up to date. As data is so central to AI, we start by explaining how training data determines the accuracy of its model.
Views: 2
