For retailers and brands still treating product data as an SEO problem, this episode is a direct challenge. Simon Dyer of Mirakl argues that the structural shift underway is not faster search but a different kind of search entirely: AI agents acting on behalf of customers, asking intent-rich questions that current product catalogues simply cannot answer. The conversation moves from marketplace economics and retail media flywheels through to a specific and actionable claim -- that community-generated data is the one truly defensible moat -- and closes on the emerging trust battle between platform giants competing to own the protocol layer of agentic commerce.

Key themes

• From SEO to GEO. As AI agents replace keyword search with conversational intent queries, product catalogue optimisation shifts from search engine optimisation to generative engine optimisation. Retailers who hold colour, size, and material data but not contextual, emotional, or situational attributes risk becoming invisible to the LLMs making recommendations on customers' behalf.
• Community data as the defensible moat. When brand product data is commoditised -- every retailer receives the same feed from New Balance -- the differentiator is proprietary community conversation: reviews, forum threads, and user-generated context that answer questions the manufacturer never thought to address. Simon's argument is that this data, structured so LLMs can find it, is where the recommendation competition will be won.
• The marketplace flywheel. Mirakl's model connects operators (retailers), sellers (brands and third parties), and customers in a self-reinforcing loop. Adding retail media to the mix creates a second revenue stream -- 70 to 80 per cent margin on promoted placements -- that scales as the seller ecosystem grows, solving the labour problem of managing hundreds of sellers through self-serve access.
• AI as the expert executioner. Simon's operating principle inside Mirakl is that AI executes faster, more completely, and more deeply than humans can, while humans define the process and make the strategic decisions. The balance he is watching for is the point at which the system has learnt his decision-making patterns well enough that he stops reviewing its choices.
• Sales reinvented: before and after. The two highest-value applications Simon describes are pre-meeting briefing (agents pulling from Salesforce, web, call recordings, and market data into a single brief) and post-meeting follow-up (summarised, multi-threaded, specific to each stakeholder). The drudgery of note-taking and CRM updating is automated; the relationship work is not.
• The protocol battle. Simon draws an explicit parallel between the current competition among Google, Apple, banks, and others to own agentic commerce infrastructure and the Betamax/VHS format war. The winner will define the data standards through which AI agents make purchases on customers' behalf. Trust -- specifically, willingness to open personal data -- is the unlock, and it remains unresolved.

⠀

What you'll learn

• Why product data optimised for keyword search fails conversational AI agents, and what GEO requires instead.
• Which type of data is genuinely proprietary to retailers in a world where brand feeds are shared universally.
• How a marketplace retail media flywheel generates margin without proportional increases in headcount.
• What a working multi-agent sales pipeline looks like in a B2B software business today, end to end.
• Why the next sales hire should already be automating parts of their personal life as a proof point of AI fluency.
• Where the trust and data-standard battles of the next 18 months are likely to be fought, and by whom.

⠀

Chapter structure

• ~00:00 Introductions: Simon Dyer and Mirakl's marketplace and drop-ship model
• ~02:00 The department store analogy: extending range without tying up capital in stock
• ~03:00 Commerce explosion: everything, everywhere, all the time as a genuine operating reality
• ~05:00 Retail media as a natural extension of the marketplace flywheel; 70--80% margin on promoted placements
• ~08:00 The structural AI shift: agents acting on customers' behalf, intent-based discovery replacing keyword search
• ~11:00 GEO versus SEO: optimising product catalogues for LLM recommendation, not search ranking
• ~14:00 Where the data value sits: brand feeds as commodity, community data as moat
• ~18:00 Simon's career arc: Siebel, Oracle, enterprise software into Mirakl
• ~19:00 AI inside Mirakl: agent-building at grassroots level, demand generation pipelines, automated CRM
• ~22:00 Brokering the AI cacophony: summarisation as the most valuable daily use of AI
• ~24:00 The expert executioner model: AI executes, human decides
• ~26:00 The next sales hire: prep, follow-up, and evidence of personal AI fluency
• ~29:00 The Betamax/VHS protocol battle: Google Universal Cart and the race to own agentic standards
• ~32:00 Trust as the limiting factor: no-quibble reliability as the foundation for autonomous purchasing

⠀

About the guest

Simon Dyer is Regional VP at Mirakl, responsible for the UK, Nordics, Middle East and Africa. Mirakl provides the platform infrastructure for retailers and brands to operate marketplace and drop-ship models, with a retail media layer built on top of the seller ecosystem. Simon's background spans enterprise software sales at Siebel and Oracle before moving to Mirakl, where he works with platform operators across retail, distribution, and digital commerce. His focus is on the commercial and structural implications of AI for marketplace economics and for the sales function specifically.

Quotes

"People aren't searching for white sneakers any more. They're saying: I've got a wedding in Italy, I'm wearing a blue suit, it'll be 35 degrees. Give me the top three and the pros and cons." 

"Retailers who aren't considering this yet will become invisible to the LLMs as they make their recommendations." 

"Community conversation -- that is where I really believe the difference will be. And if that data is structured in a way LLMs can find it, you've kind of won the recommendation competition." 

"AI is the expert executioner. The human is the strategic thinker."

"Your imagination is the only thing holding you back right now in how to use these tools."