Hey readers make sure to check out the new tools and modules I added to https://valueaddvc.com

You can search 1000+ VCs here https://valueaddvc.com/vc-universe

Or match like a dating app here with them https://valueaddvc.com/lp-match

There’s a growing narrative that AI agents plus stablecoins are about to eliminate interchange and route around the card networks. The argument is straightforward. Models are improving. Agents can transact. Stablecoins settle instantly. Remove humans and friction disappears.

I don’t think it plays out that way.

I say that not only as someone who invests in AI, but as someone who spent years at American Express working on B2B payments infrastructure. I was on the inside of a major network trying to modernize pieces of it. Payments look simple from the outside. They are not.

They are not primarily a routing problem. They are a trust, liability, regulatory, and coordination problem.

Let’s unpack it.

Yes, models are improving rapidly. Anthropic and others are building agents that can manage workflows and potentially execute transactions.

That matters at the application layer.

But intelligence does not replace the underlying system that makes a payment legally valid and financially final. That system includes:

• Issuer licensing
• AML enforcement
• Fraud monitoring and loss reserves
• Chargeback frameworks
• Merchant underwriting
• Cross border compliance
• Regulatory reporting

Inside American Express, even modest changes to payment flows required coordination across risk, legal, compliance, banking partners, and regulators. Moving money is not just about API calls. It is about capital requirements, supervision, and defined accountability.

An agent can initiate a transaction. It cannot by itself solve the institutional coordination required to make that transaction compliant at global scale.

When consumers swipe a card, they are not just buying convenience. They are buying liability protection.

If something goes wrong:

• The issuer reverses the charge
• Fraud losses are absorbed
• Disputes are adjudicated
• Legal responsibility is clearly assigned

That allocation of loss is the product.

If an AI agent makes an unauthorized purchase or misinterprets instructions, who is liable?

• The model provider
• The wallet custodian
• The stablecoin issuer
• The merchant
• The end user

Until liability is clearly defined, regulated, and consistently enforced, agents cannot replace cards at scale. Payments ultimately come down to who absorbs the loss.

That question must be answered before consumer retail can migrate.

Retail payments are among the most regulated systems in finance. Any alternative must address:

• Licensing across jurisdictions
• Identity verification
• AML and KYC standards
• Consumer protection rules
• Data privacy laws
• Cross border legal alignment

This requires coordination across banks, regulators, and governments. It is not something that moves at startup speed.

From experience, regulatory alignment is often the gating factor. It shapes product design more than engineering ambition does.

Interchange is often described as pure margin.

In practice, it funds:

• Fraud detection systems
• Dispute resolution mechanisms
• Network security
• Compliance operations
• Risk underwriting

Visa and Mastercard process tens of trillions annually across millions of merchants and thousands of financial institutions. That infrastructure supports uptime, fraud coverage, and legal clarity at scale.

If stablecoin rails expand into mainstream retail commerce, they will need to rebuild many of these same functions. When that happens, cost structures return. The economics do not simply compress to zero.

Most major stablecoin issuers monetize primarily through interest income on reserves held in short duration government securities. That model is rate sensitive.

Card networks monetize:

• Payment volume
• Cross border flows
• Data services
• Fraud tooling
• Long term network relationships

These are different economic foundations. A treasury backed token is not equivalent to a global switching network embedded in banking systems.

Stablecoin transaction volumes can appear very large.

Much of that activity is crypto native. Exchange settlement, arbitrage, and treasury movements between trading entities.

That is not the same as regulated retail commerce across physical and digital merchants, with consumer protection and dispute rights attached.

Processing global consumer payments at scale requires more than settlement speed. It requires regulatory compliance, defined liability, and operational resilience.

Consumers use cards because they receive:

• Points
• Cashback
• Travel benefits
• Purchase protection
• Fraud coverage

An AI agent optimizing merchant fees does not automatically align with consumer incentives.

Most consumers are not going to manage wallets, custody keys, or tax implications to reduce a cost they do not directly perceive. Adoption follows incentives and habit, not theoretical efficiency.

The more plausible outcome is integration.

• Stablecoins may serve as additional settlement rails
• Agents may automate B2B and cross border flows
• Networks may incorporate tokenized infrastructure under existing consumer interfaces

Retail payments are governed by trust, liability clarity, regulation, and deeply embedded infrastructure. Those systems evolved over decades.

AI will influence payments. It may meaningfully reshape parts of the stack.

But replacing global consumer card networks requires rebuilding the legal, regulatory, and risk architecture underneath them. That is a structural challenge, not an API problem.

I drafted the original version of this myself, then used AI to refine it. That feels like the right model. Intelligence enhances systems. It does not automatically replace them.

While you were debating if AI would take your job, other people started using it to print money. Seriously.

That's not hyperbole. People are literally using ChatGPT to write Etsy descriptions that convert 3x better. Claude to build entire SaaS products without coding. Midjourney to create designs clients pay thousands for.

The Hustle found 200+ ways regular humans are turning AI into income. Subscribe to The Hustle for the full guide and unlock daily business intel that's actually interesting.

Start making AI money