Inside Heineken’s Flavour Innovation: How Agentic Ideation and AI Personas Are Boosting Novelty

At the 2025 Pangborn Sensory Symposium, MMR Research Worldwide’s Alexandra Kuzmina was awarded Best Oral Presentation in the Sensometric session for a pioneering collaboration with Heineken. The project introduced the Agentic Ideation Pipeline (AgIP) – a novel system designed to enhance the novelty of AI-generated product and flavour ideas by drawing on internal and external data sources.

In today’s competitive consumer landscape, brands face a paradox: they collect vast amounts of data, yet breakthrough innovation remains elusive. Incremental improvements are common, but truly novel product and flavour ideas, those that surprise and delight, are rare. The challenge is not just in gathering insights, but in connecting disparate data points to spark creativity.

Heineken partnered with MMR to push the boundaries of AI-driven ideation. Building on the success of AI personas, which are chat-based interfaces grounded in real consumer segmentation data from Heineken’s flavour segmentation research, the team developed the Agentic Ideation Pipeline. Inspired by recent academic research in other domains, this multi-stage, iterative system was designed to enhance the novelty and diversity of new product and flavour ideas.

1. Initial Seed Idea Generation:
   The system uses a multi-source knowledge base to generate seed ideas from both internal (Heineken’s flavour segmentation and proprietary consumer research & knowledge database) and external (MMR’s brand agnostic knowledge & thought leadership database) data.
2. Seed Idea Improvement:
   Each seed idea enters an enhancement loop, where AI agents retrieve knowledge and refine the ideas over multiple cycles, increasing novelty.
3. Idea Completion:
   After multiple iterations, the most promising ideas that best represent the range and potential of the seed idea space are identified and given as a final output.

To test the hypothesis that iterative enrichment enhances novelty, identical prompts were given to both the current Heineken’s AI personas tool and the Agentic Ideation Pipeline. Each generated 10 new flavour ideas, which were randomized and blinded for evaluation by 354 UK consumers and 6 Heineken stakeholders.

• Novelty:
  Ideas from the Agentic Ideation Pipeline were perceived as more novel by both consumers and stakeholders. These ideas occupied a distinct emotional territory, like ‘adventurous’, ‘daring’, and ‘intriguing’, compared to the more conventional outputs of the standard AI personas.

• Appeal and Purchase Intent:
  While both sets of ideas were appealing, the Agentic Ideation Pipeline’s outputs required greater cognitive effort to form an opinion, resulting in a niche appeal. However, segmentation revealed a sizeable group of consumers who strongly preferred the novel ideas, indicating a clear market for innovation.

• Strategic Implications:
  Brands can use this approach to tailor innovation and marketing strategies to specific segments seeking novelty, optimizing product ranges and differentiating themselves in the market.

Heineken has long focused on bringing consumer closeness into its innovation process, and AI personas made that intuitive. But the Agentic Ideation Pipeline took things further because it showed how all the data already available could be used in a more meaningful and powerful way.

The outputs weren’t just new: they were measurably more novel and emotionally distinctive. Crucially, they resonated with a specific segment of consumers actively seeking something different, opening up clear opportunities to tailor innovation and marketing strategies.

This approach demonstrates how AI personas can be used not just to reflect consumer understanding, but to actively drive innovation, unlocking richer, more varied insights that fuel the next wave of product and flavour development.