Thanks for the write up, Casey. I'm reminded of something I wrote internally for TeePublic (an artist-apparel marketplace) way back in 2020 when Google came out with "shopping actions" and Facebook came out with "shop"...
We are at risk of being turned into an undifferentiated commodity: Google, Facebook, and Amazon are fighting a war over who owns our customers, offering users the ability to never leave their site in order to purchase our products. The issue is compounded by the fact that marketing has operated as a commodity on these platforms and reinforced this problem. Google shopping actions allow a user to search for and purchase a product without leaving Google. We cannot re-market to these users, since we are only a "fulfiller" (a nameless, faceless Google seller). We run the risk of being "Amazoned" i.e. when people ask where they got their t-shirt they respond “I got it on Google."
I think this points to your emphasis on "fortifying brand hooks with customers via the website experience post the handoff from the LLM." This has always been a problem for marketplaces: when you don't own the relationship, you're just a backend API for someone else's UX.
What got us out from underneath Google/Facebook was a heavy investment in brand, product, and the post-click experience that gave people a reason to remember us.
Yep. Totally agree. I think Facebook Shops / Tiktok Shops felt less inevitable than agent world. Deciding to sell on Amazon and integrating inventory with Google Shopping ended up being the bigger Faustian bargains.
If your product roadmap for 2025 doesn't include "How do we compete with/become an AI agent," you're planning for yesterday's game.
This isn't just a marketplace problem—it's a PM competency problem. The best product minds are already repositioning their entire value props around this shift.
Shoutout to Casey Winters for connecting dots most of us didn't even see. This is the kind of strategic thinking that separates good PMs from great ones.
Great piece. Couple quick supporting observations from the travel industry:
* I think it's no accident that the CEO Expedia named after ChatGPT launch was Adrian Gorins, who came up through the B2B side of the business (Expedia Affiliate Network, Expedia for Business). Non consumer direct channels become increasingly important.
* Most travel marketers (at least for multi-tenanted inventory like hotels) have shifted to generate positive contribution margin on a first transaction basis. They already know there's little customer value to be had.
* Maybe one more claim to make on how marketplaces can respond: differentiate on the consumer experience. Booking.com has done with masterfully with their Genius program (providing up to 20% price discounts for loyal customers). This has allowed Booking to significantly reduce marketing as a % of revenue over the past 2 years and drive more direct traffic. And in an agent-based world, if Booking can provide sustainably lower prices for high frequency travelers, that's a value prop that forces LLM-based agents to use them / consumers to continue to buy direct. More broadly, can marketplaces use their scale to create diffrentiated experiences and lower prices for loyal customers?
Certainly easier for incumbents than startups to adapt on both on the customer acquisition side and maybe things like pricing. Radically new product experiences favor startups of course, and that could lead to M&A from incumbent marketplaces.
love how deep you go casey. both-siding agents is great. we're trying to build agents to intercept agents, from the supply side. this is going to get fun. and dirty!
“Fortify brand hooks with customers via the website experience post the handoff from the LLM”
If the initial sale came in through an agent, wouldn’t the site fortify its “brand” hook with the LLM agent instead? The after-sale follow up is all about smoothing the path to repeat purchase which might also be agent-driven
We've built agents for SuperMe on the supply side, but it's a proto-marketplace at this point.
I didn't include examples since the ones I've seen are so nascent I'm not sure they are impactful. Amazon has Rufus. Doordash and Faire have tried some things, but can't tell if they work yet.
Thanks for the write up, Casey. I'm reminded of something I wrote internally for TeePublic (an artist-apparel marketplace) way back in 2020 when Google came out with "shopping actions" and Facebook came out with "shop"...
We are at risk of being turned into an undifferentiated commodity: Google, Facebook, and Amazon are fighting a war over who owns our customers, offering users the ability to never leave their site in order to purchase our products. The issue is compounded by the fact that marketing has operated as a commodity on these platforms and reinforced this problem. Google shopping actions allow a user to search for and purchase a product without leaving Google. We cannot re-market to these users, since we are only a "fulfiller" (a nameless, faceless Google seller). We run the risk of being "Amazoned" i.e. when people ask where they got their t-shirt they respond “I got it on Google."
I think this points to your emphasis on "fortifying brand hooks with customers via the website experience post the handoff from the LLM." This has always been a problem for marketplaces: when you don't own the relationship, you're just a backend API for someone else's UX.
What got us out from underneath Google/Facebook was a heavy investment in brand, product, and the post-click experience that gave people a reason to remember us.
Yep. Totally agree. I think Facebook Shops / Tiktok Shops felt less inevitable than agent world. Deciding to sell on Amazon and integrating inventory with Google Shopping ended up being the bigger Faustian bargains.
If your product roadmap for 2025 doesn't include "How do we compete with/become an AI agent," you're planning for yesterday's game.
This isn't just a marketplace problem—it's a PM competency problem. The best product minds are already repositioning their entire value props around this shift.
Shoutout to Casey Winters for connecting dots most of us didn't even see. This is the kind of strategic thinking that separates good PMs from great ones.
Great piece. Couple quick supporting observations from the travel industry:
* I think it's no accident that the CEO Expedia named after ChatGPT launch was Adrian Gorins, who came up through the B2B side of the business (Expedia Affiliate Network, Expedia for Business). Non consumer direct channels become increasingly important.
* Most travel marketers (at least for multi-tenanted inventory like hotels) have shifted to generate positive contribution margin on a first transaction basis. They already know there's little customer value to be had.
* Maybe one more claim to make on how marketplaces can respond: differentiate on the consumer experience. Booking.com has done with masterfully with their Genius program (providing up to 20% price discounts for loyal customers). This has allowed Booking to significantly reduce marketing as a % of revenue over the past 2 years and drive more direct traffic. And in an agent-based world, if Booking can provide sustainably lower prices for high frequency travelers, that's a value prop that forces LLM-based agents to use them / consumers to continue to buy direct. More broadly, can marketplaces use their scale to create diffrentiated experiences and lower prices for loyal customers?
Certainly easier for incumbents than startups to adapt on both on the customer acquisition side and maybe things like pricing. Radically new product experiences favor startups of course, and that could lead to M&A from incumbent marketplaces.
The discovery orchestration, plus switching the marketplace paradigm directly onto the AI rails or agents themselves feels like the future
love how deep you go casey. both-siding agents is great. we're trying to build agents to intercept agents, from the supply side. this is going to get fun. and dirty!
“Fortify brand hooks with customers via the website experience post the handoff from the LLM”
If the initial sale came in through an agent, wouldn’t the site fortify its “brand” hook with the LLM agent instead? The after-sale follow up is all about smoothing the path to repeat purchase which might also be agent-driven
A2SPA secures agents at a protocol level which changes the game on agent deployments.
Thanks for sharing the thoughtful write-up, Casey.
> Go agentic on both sides of the marketplace. Develop your supply and demand agents and capture your own structured data. Use AI to beat AI.
Have you seen any examples of marketplaces that are doing this effectively?
We've built agents for SuperMe on the supply side, but it's a proto-marketplace at this point.
I didn't include examples since the ones I've seen are so nascent I'm not sure they are impactful. Amazon has Rufus. Doordash and Faire have tried some things, but can't tell if they work yet.
Got it, feels like it's super early days still. If you're ever up for swapping notes, I'd love to chat (also building an "AI-native" marketplace).