Innovation has never moved faster. Tasks that once took weeks, such as generating concepts, analysing trends and creative development, can now happen in hours. Innovation teams can explore hundreds of ideas instead of a handful, making iteration much faster.
So why isn’t that translating into better products? Industry data suggests that two in three innovations are failing by year two, meaning that only a minority of ideas that make it to the shelf are hitting the mark.
This doesn’t just mean missing out on sales; it has implications for the brand as well. Innovation is one of the most important drivers of brand growth. Research shows that brands that are both meaningfully different to their competitors and innovative are twice as likely to grow as brands that are not.
So why are innovation teams struggling to turn all this stimulus into products that people love and want to return to? Food and drink brands are surrounded by fragmented and often contradictory signals from consumers. Are people prioritising treating themselves, proteinmaxxing, avoiding UPFs, or some combination of all the above?
A few years ago, innovation teams could rely on a relatively small number of inputs like surveys, focus groups and sales data. Now these insights into what people want are scattered across multiple platforms and behaviours rather than coming from a few clear sources.
This makes it easier for brands to get swept up in the latest hype cycle, whether or not it’s a good fit for the brand. That’s how we end up with shelves flooded with products that are all variations on the same theme: Dubai chocolate and ube flavoured everything. The result is that teams are left wondering why they failed to spark real consumer interest.
The problem with LLMs
In this environment, the rise of AI is a double-edged sword. LLMs might be a natural starting point for exploring ideas, but leaning on ‘off the rack’ AI tools will only create a recipe for blandification.
LLMs can certainly come up with a lot of ideas, but much of its output is generic and disconnected from the deeper motivations behind consumer behaviour. Traditional large language models (LLMs) also tend to flatten nuance - producing quantity over quality.
When Kantar analysed the outputs of three of the leading LLMs, we found that only four in ten of the opportunities and ideas they generated could be classed as ‘strong’ or ‘best in class’. In the same way that using ChatGPT to write a thank you note will give you something that sounds polished, but not very personal, using AI tools without the right inputs will just scale sameness.
For NPD that drives growth, you don’t just need more ideas – you need the right ones. Ideas that are rooted in real human behaviour and clear about the role the brand can play within that. To achieve that you need to blend the best of AI and human intelligence to filter out noise and focus on the ideas that will genuinely drive growth.
The good news is that the next generation of AI-enabled innovation tools are already demonstrating what is possible when technology is paired with a deeper understanding of people, brands and culture. But leaning on generic AI tools to generate ideas will end up costing teams time spent chasing trends, without the depth of understanding needed to identify opportunities that truly matter.
About the author
Dr Nicki Morley is the global MD of innovation solutions at Kantar.
