Can AI facilitate market research and innovation?

Generative artificial intelligence (GenAI) is reshaping how businesses identify consumer needs, develop products and services, and design packaging and communication. Tools such as ChatGPT, Grok, Gemini, Perplexity, and Claude now support insight generation, accelerate experimentation, and reduce cost and time to market.

In sensory and consumer research, which examines how people experience products such as foods and beverages through their senses, GenAI is beginning to impact how ideas are generated, tested, and commercialised. Firms experiment with it to develop flavours, packaging, and market forecasts. Sony, for instance, built interactive AI prototypes to inspire recipe creation, Symrise developed Symvision AI™ to analyse flavour trends, and Packify.ai generates early packaging visuals.

GenAI can operate across three iterative phases, concept, design, and testing, each refined through successive, iterative, prompting. However, dependence on GenAI entails risks, including training-set bias, privacy concerns, and simplified representations of consumer behaviour. Effective adoption requires combining structured prompting with expert human judgement, treating AI output as input for optimisation rather than as final evidence.

Concept: Sparking creativity and hypotheses

Innovation begins with concept generation. In 2024, visitors at Johnnie Walker Princes Street co-created Blue Label packaging using three prompts that produced a design in Scott Naismith’s style within minutes. More broadly, text-based tools such as ChatGPT or Gemini, and image generators like MidJourney, now serve as creative partners, producing descriptions and visual mock-ups. For instance, a beverage firm might test prompts such as: Would slimmer, translucent bottles appear healthier and more premium? Could metallic finishes enhance freshness cues? Prompt engineering—adjusting wording, adding constraints, or referencing specific styles—refines such outputs. Effective prompting in sensory and consumer research involves:

• Clarity: Use precise wording and context (e.g., “premium sparkling water packaging for health-conscious consumers”).
• Constraints: Set boundaries for style, material, or sustainability.
• Anchoring: Link prompts to consumer psychology or market data.
• Iteration: Refine prompts incrementally using feedback.
• Documentation: Record prompts and outputs for transparency and replication.

Indeed, effective concept development requires data grounding. Firms increasingly combine GenAI with consumer psychology, sensory mapping, trend databases, or consumer data to ensure ideas are both imaginative and evidence-based. Evaluation criteria such as differentiation, feasibility, and brand alignment then guide which AI-generated concepts advance to design.

Design: Generating and validating ideas

Turning concepts into testable designs can be costly and time-consuming; GenAI reduces this substantially. Text-to-image tools such as MidJourney or DALL·E 2 can generate bottle shapes, textures, or metaphor-driven packs in minutes. Platforms like Grok’s Imagine extend this to short videos, accelerating design cycles and lowering costs.

Coca-Cola used GenAI mood boards to develop Y3000’s futuristic packaging, and Mattel adopted Adobe Firefly to speed up toy design. Prompt engineering enables adaptation to regional markets when AI systems are guided to incorporate local languages, cultural nuances, sensory preferences, and contextual factors that shape consumer interpretation and response: a firm developing plant-based snacks might generate culturally tailored visual identities and survey materials in multiple languages.

A key challenge is translating AI visuals into manufacturable products. Integration with design software and packaging simulators helps test material tolerances, sustainability metrics, and logistics early. In such co-design processes, AI provides initial blueprints refined through expert input, bridging inspiration and production readiness.

Testing: From silicon samples to consumer insights

Traditional consumer testing is slow and expensive. GenAI offers an alternative via “silicon samples”: synthetic respondents simulating human responses. Large language models can reproduce sensory associations, such as linking rounded shapes and red hues with sweetness. Studies, including our own, show synthetic panels reliably capture such patterns, enabling early-stage screening before large-scale testing.

Firms can also fine-tune AI models with proprietary data for more specific insights. Yet synthetic testing should complement, not replace, human studies. Hybrid approaches, such as AI screening broad concept sets followed by targeted human validation, can optimise efficiency while preserving ecological validity. Linking GenAI-derived data to key indicators such as liking or willingness to pay, and embedding them in dashboards, ensures transparent decision criteria from concept to launch.

Implementation: Closing the loop

Once concepts are validated, GenAI can assist in scaling production, crafting communication materials, and monitoring market reception. Natural language models can aid in analysing consumer reviews, social media data, and sensory feedback to detect emerging issues or opportunities. Insights from this stage can feed back into concept prompting, forming a continuous learning cycle that refines both AI models and innovation pipelines.

Pitfalls and responsible use

Despite its promise, GenAI presents significant challenges. Training data may embed group, cultural, or political bias, shaping outputs in unintended ways. Sensitive inputs or generated content can also breach privacy standards and conflict with regulations such as the GDPR. Moreover, current models tend to oversimplify human experience, struggling to capture individual differences, which can result in distorted or misleading interpretations of consumer behaviour.

Effective governance of GenAI requires more than technical competence. Organisations should ensure transparency by documenting prompts, datasets, and model versions. Validation through expert review must complement AI-generated insights to maintain interpretative depth and contextual accuracy. Systematic auditing should be applied to detect and mitigate bias, while robust privacy measures must ensure full compliance with data protection legislation. Finally, clear disclosure of how AI has been used in product development and marketing supports accountability and consumer trust.

The bottom line

Viewing GenAI as a means of human augmentation clarifies its value: it extends, rather than replaces, human capability. The most effective use emerges when analytical rigour, ethical awareness, and creativity converge. AI can catalyse innovation, but human insight remains essential for accountability and progress.