Agentic commerce is moving quickly from demos to real-world experimentation. But most commerce infrastructure wasn’t designed for AI agents. In our new paper, “Agentic Commerce Has an Infrastructure Problem,” we outline the key constraints shaping this new model of commerce — and what they mean for merchants preparing for agent-driven transactions. 📄 Read the full paper: https://lnkd.in/e6zmuHm4
Really interesting paper. It made me think about how much current commerce flows rely on a human being in the loop to smooth over inconsistencies. Once agents start querying pricing, inventory, or configuration repeatedly over longer sessions, those gaps become much more visible. At that point the challenge in agentic commerce feels less about the agent itself and more about whether the underlying systems can support that kind of interaction pattern reliably at scale.
Most merchants are not ready for this shift. Agentic commerce doesn't browse. It decides, instantly. → Bad product data? Rejected. → Stale inventory? Transaction failed. → Inconsistent pricing? Agent moves on. No second chances. No retargeting. No abandoned cart email. The infrastructure you built for human shoppers won't work for AI agents. 2025-2026 is the window. Early movers will compound this advantage for years. Great paper, what do you see as the #1 fix merchants should prioritize first?
Very interesting perspective. As AI agents begin initiating transactions on behalf of users, it seems the challenge is not only infrastructure but also authorization boundaries, fraud detection, and liability frameworks. Distinguishing legitimate agent activity from malicious automation will likely become an important capability for payment and risk systems as agentic commerce scales.
Interesting point. Agentic commerce highlights how much of today’s commerce infrastructure was built around human interactions and explicit user flows. As agents start initiating transactions, questions around identity, authorization, trust signals, and liability will likely become core infrastructure challenges rather than edge cases.
Agentic commerce is definitely moving faster than most infrastructure can handle. Merchants that start preparing now will likely have a strong advantage. Curious to see how platforms adapt to support this new model.
Exactly. For vertical SaaS payments, the agentic shift is a power struggle over the transaction mandate. The priority right now is moving fast to own that mandate. Owning the payment rails is secondary to owning the permission to spend. That is where the real value lies in an agentic SaaS world.
Really compelling read about the shift from human-centric to agent-driven commerce.
Airlines. Are you reading along?
Plait & Pattern•2 mil seguidores
1 mConstraint #2 is the one that deserves more attention. The product data problem isn't new, it's just been invisible. Web UIs are remarkably good at hiding bad data. A missing attribute, an inconsistent unit of measure, a product weight stored in a different system than the rest of the catalog — humans route around these gaps instinctively. Agents can't. What Adyen is describing as an agentic commerce readiness problem is really a foundational data architecture problem that's been accumulating for years. The four-stage journey in this paper is a useful frame, but most enterprises aren't stuck at "does the data exist?" They're stuck in the messy middle. The data exists, partially, across multiple systems, with different update cadences, inconsistent attribute definitions, and no single source of truth. Getting from "data exists somewhere" to "data is consistent and machine-queryable" is where most of the work, and most of the value, actually sits. And the companies that fix this now aren't just preparing for agents. They're fixing what's already costing them in search, in conversion, and in returns.