What breaks when AI agents do the shopping? (200-page book)
3 comments
References I’m pointing to (so you don’t have to take my word for it):
- Google: Universal Commerce Protocol + “agentic commerce” checkout surfaces https://blog.google/products/ads-commerce/agentic-commerce-a... https://developers.googleblog.com/under-the-hood-universal-c... - Shopify engineering write-up on UCP https://shopify.engineering/ucp - OpenAI: “Buy it in ChatGPT” https://openai.com/index/buy-it-in-chatgpt/ - Amazon: “Buy for Me” (beta) https://www.aboutamazon.com/news/retail/amazon-shopping-app-... - Visa: secure agentic transactions https://corporate.visa.com/en/sites/visa-perspectives/newsro... - Mastercard: Agent Pay / agentic payments https://www.mastercard.com/us/en/news-and-trends/press/2025/...
- Google: Universal Commerce Protocol + “agentic commerce” checkout surfaces https://blog.google/products/ads-commerce/agentic-commerce-a... https://developers.googleblog.com/under-the-hood-universal-c... - Shopify engineering write-up on UCP https://shopify.engineering/ucp - OpenAI: “Buy it in ChatGPT” https://openai.com/index/buy-it-in-chatgpt/ - Amazon: “Buy for Me” (beta) https://www.aboutamazon.com/news/retail/amazon-shopping-app-... - Visa: secure agentic transactions https://corporate.visa.com/en/sites/visa-perspectives/newsro... - Mastercard: Agent Pay / agentic payments https://www.mastercard.com/us/en/news-and-trends/press/2025/...
One page summary of the book:
The “agentic commerce” stack: 1) Intent (what the user actually needs) 2) Research (what options exist) 3) Evaluation (tradeoffs + trust signals) 4) Transaction (permissions/auth/payment) 5) Fulfillment (delivery, returns, satisfaction loop)
What changes vs today: - SEO/ads matter less if agents don’t browse pages - structured product data + availability + policies matter more - reputation signals become machine-readable and portable - checkout becomes delegated + constrained (limits, identity, auditability)
If you want to argue against it, the best angles are: - why humans will keep browsing (and for which categories) - where trust signals will come from (and who controls them) - whether retailers will allow agents at scale
The “agentic commerce” stack: 1) Intent (what the user actually needs) 2) Research (what options exist) 3) Evaluation (tradeoffs + trust signals) 4) Transaction (permissions/auth/payment) 5) Fulfillment (delivery, returns, satisfaction loop)
What changes vs today: - SEO/ads matter less if agents don’t browse pages - structured product data + availability + policies matter more - reputation signals become machine-readable and portable - checkout becomes delegated + constrained (limits, identity, auditability)
If you want to argue against it, the best angles are: - why humans will keep browsing (and for which categories) - where trust signals will come from (and who controls them) - whether retailers will allow agents at scale
I lead an Amazon agency and wrote a 200+ page book about a shift I think is already underway: when the buyer is an AI agent, the “shopping cart” becomes mostly invisible.
My core model: users set policies (budget, brands to avoid, delivery constraints, quality threshold). An agent does discovery + evaluation + checkout inside those constraints.
This changes how products win: - discovery shifts from SEO/ads to structured product data + reputation signals - “conversion” becomes permissioning + delegated payment + trust - brands start optimizing for agent evaluation, not human browsing
I’d love pushback from anyone building in delegated auth, product data standards, or trust/reputation. What’s overstated / missing here?