Where we are in 2025
Here are agentic commerce teardowns, based only on observable behaviour, mapped to the lifecycle model.
1. Instacart AI (Shopping Assistant / Recipes → Cart)
Instacart’s AI helps users:
- plan meals
- suggest recipes
- add items to cart
- modify quantities
Product Lifecycle mapping (observable)
The AI can:
The AI cannot:
Lesson:
Cart assembly is not commitment (AI allowed). Payment is. (No AI)
Instacart treats the cart as a reversible buffer, not a commit slice.
2. Shopify Magic / Sidekick (Merchant-side commerce)
Shopify AI helps users:
- write product descriptions
- analyse store performance
- suggest actions
Product Lifecycle mapping (observable)
Shopify AI can:
Shopify AI cannot:
Lesson:
Merchant-facing AI avoids commit authority by design.
This isn’t lack of ambition, it’s risk management.
3. Klarna AI Shopping Assistant
Klarna’s assistant helps users:
- find products
- compare prices
- manage purchases and payments
Product Lifecycle mapping (observable)
Lesson:
Klarna operates close to money, yet still:
- requires explicit confirmation
- maintains UI-led control points
Even fintech-grade AI keeps a strict commit boundary.
Proximity to payment ≠authority over payment.
4. Expedia / Booking.com AI Trip Planners
What they do:
- suggest itineraries
- compare flights and hotels
- answer “what’s best” questions
Product Lifecycle mapping (observable)
Lesson:
Even though booking could be automated, it isn’t. Because:
- cancellations
- policies
- dates
- legal terms
The more complex the commitment, the more explicit the handoff to humans.
Comments are closed.