AIBRIEF
AI shopping agents need line by line receipt data to apply offers and pick products. That same detail is also what lets the network behind the agent see exactly what you bought.
An AI shopping agent doesn't get far on a card number alone. To pick between two near-identical yogurts, apply the right coupon, or push the right loyalty perk, it needs the line-item detail on the digital receipt, not just the merchant and total.
That distinction is the load-bearing idea in a recent PYMNTS opinion piece by Mladen Vladic, head of product management for payment networks at FIS. Transaction-level signals (where the card swiped, how much it cost) have been around for decades. Item-level intelligence is newer: a complete, real-time view of what was actually bought, down to the SKU.
Vladic anchors the case on McKinsey's October 2025 projection of up to $1 trillion in U.S. agentic-commerce revenue by 2030, with $3 trillion to $5 trillion globally. Those are forward-looking figures, not measured sales, and worth treating as a vendor-cited projection rather than a confirmed market.
The same data that lets an agent act also lets its vendor stack profile the buyer. FIS positions its Smart Basket product as one answer; the article lists three guardrails (approved payment methods, existing auth networks, spending limits held by the account holder). What it does not name is a non-vendor voice. A regulator, merchant, or independent analyst has not yet tested those guardrails against practice. That is the lane still open.