James Taylor, Founder and CEO of Particular Audience, joined the Innovation Focus panel at the dunnhumby Retail Innovation Forum in London — alongside leaders from dunnhumby, Reward, Diageo, Provenance and Lloyds Banking Group. Panel: Innovation focus: what do consumers want? — A no-slides, real-signal discussion on consumer expectations, lean iterative testing and retail media.

Panellists: James Taylor (Particular Audience), Gaby Appleton (dunnhumby), Darren Wright (Reward), Benjamin Lickfett (Diageo), Sarah Arana-Morton (Provenance), Paul Wilkinson (Lloyds Banking Group).

Test and learn with AI: Using AI to quickly test and learn what consumers actually want — lean, iterative and grounded in real signals.

Tighter margins, new expectations: Navigating tighter margins and shifting expectations of what retail media should deliver for brands and retailers.

Transparent partnerships: Collaborative partnerships and transparent communication frameworks across brands and retailers.

Value sets the boundary, but values still decide. Shoppers are more cost-conscious than two years ago. Within price boundaries, health, sustainability, quality and convenience act as tie-breakers — and shoppers expect credibility, not slogans. What this means: Win on price credibility first, then use trustworthy product data to differentiate on values.

Agentic commerce is real, but filtering comes before full delegation. Traffic to Gen-AI browsers has reportedly grown more than 5,000% in the US (as reported by dunnhumby). Shoppers let AI narrow choices but don't yet hand over the purchase — and hallucinated prices, returns or availability become a financial risk. Open standards like Model Context Protocol make deep integration far cheaper. What this means: Invest in AI-ready infrastructure now: accuracy, real-time data flows and open protocols over flashy front ends.

Proof beats persuasion. As discovery moves through AI, product data becomes the brand. Verified claims and structured, machine-readable proof make brands visible to agents; persuasion alone gets filtered out. What this means: Treat product data as a strategic asset — audit claims and structure proof.

Personalisation must earn the data it uses. Recency and real-time context now matter more than long historical profiles. The product being viewed, the basket being built and the question being asked predict intent better than persona data — the bigger risk is under-personalisation. What this means: Move from profile-based to context-based personalisation.

Loyalty is a repertoire, not a relationship. Even loyal shoppers buy from a repertoire of about seven brands, and the mix shifts every year. AI agents will accelerate fragmentation — loyalty is earned in every interaction. What this means: Compete for repeat selection, not exclusivity.

Innovation discipline beats innovation volume. Prototypes take an hour and synthetic users can stress-test concepts. The constraint is no longer tools — it's strategic focus and measurement discipline. What this means: Cut the experimentation portfolio, sharpen measurement and pre-build execution plans.

Takeaways summarised from What do consumers really want? Trust, value, and the shape of AI-driven retail — dunnhumby, 24 July 2026 (Leo Nagdas, Nate Kovarik). Quotes are from the panel discussion and are not attributed to individual panellists.

What do consumers really want? Trust, value, and the shape of AI-driven retail · Agentic Commerce & MCP · Personalisation