AI Agents Are Coming for Commerce — And Big Tech Says Crypto Rails Are the Answer

"An agent cannot get a bank account."

Richard Widmann, global head of Web3 strategy at Google Cloud, didn't sugarcoat it on stage:

"An agent cannot get a bank account. It's not hard, it just is impossible."

He pointed at both technological and regulatory barriers, then described crypto as "a fantastic machine readable interface for payments."

That single line reframes the conversation. Crypto isn't being positioned as an alternative to banking. It's being positioned as the only payments layer agents can actually use.

Google's AP2 + 120 Partners

To address the gap, Google launched the Agentic Payments Protocol (AP2) — an open protocol donated to the FIDO Foundation with more than 120 partners, PayPal among them. Widmann compared the move to x402, the internet-native payment standard given to the Linux Foundation.

"Open dialogues and open standards are really the foundation you need to build on."

PayPal: Agents Are the Next Channel

May Zabaneh, SVP and GM of crypto at PayPal, framed agents as the next channel after offline → online → mobile. PYUSD, she said, is "a very natural programmable layer for payments" — particularly as commerce trends toward globalization, AI-native experiences, and tokenized assets.

The data she shared lands hard. A recent PayPal survey found:

• 95% of merchants now see AI agent traffic on their sites
• Only 20% have machine-readable catalogs

"Merchants need to be ready for this next era."

The shift, she added, mirrors the move from offline to online stores. Merchants need to expose their products in agent-readable formats — or get skipped.

Liability, Custody, and the Trust Problem

Asked who's responsible if an agent makes a bad purchase, Zabaneh called it "definitely something that we have to think through as an industry."

Widmann said multi-party custody is becoming central to agent design. Google has extended its Cloud KMS platform to crypto custody, and he argued agents should hold only one of two or three key shards rather than full keys:

"It cannot simply unilaterally move funds or take action."

Translation: as soon as agents start spending real money, the entire stack needs verifiable controls, audit trails, and proof of who authorized what. That's not a payments problem. That's a trust infrastructure problem.

This Is Where the $DAG Thesis Gets Loud

AI agents won't just chat. They'll buy, negotiate, verify, settle, and interact with the real world on their own. That creates one massive problem:

How do you trust autonomous systems at scale?

Constellation Network has spent years quietly building exactly the layer that question demands:

• Verifiable data
• Digital Evidence
• Tamper-proof audit trails
• Real-world data validation
• Enterprise-grade trust infrastructure

As AI evolves from generating answers → to taking actions, trust becomes the product. And trust needs infrastructure.

What Keeps Them Up at Night

Asked what keeps him up at night, Widmann said the open question is "how do you onboard agents into all of the existing capital markets and infrastructure plumbing that powers payments and trading today."

Zabaneh said trust is what keeps her up professionally — though personally, she "can't wait for agentic to help make my life easier."