Will AI Swipe - or Tap Smart Contracts Instead?
Visa and Mastercard now let AI agents spend, but smart contracts may be the rails AI truly prefers.
This week, Visa and Mastercard stepped into the AI spotlight with back-to-back announcements Visa’s “Intelligent Commerce” and Mastercard’s “Agent Pay.”
It follows PayPal’s move during Dubai AI Week to launch merchant-facing AI agents. If you missed it, catch up here.
AI agents can now spend money. Think: ask ChatGPT to find your ideal Airbnb for your next trip, and when you’re ready, say, “book it.” The agent handles payment on your behalf. Not bad.
In this article, we’ll break down:
💳 Visa and Mastercard shipped - and what’s just marketing gloss.
➡️ Why smart contracts are the more natural fit for AI spending agents.
🔎 How AI’s relentless cost optimisation puts every payment rail under pressure.
📱 What cards still do better (for now), and where they’re vulnerable.
Hey there, Dom here 👋
A quick note to welcome all the new subscribers - I still have to pinch myself each week seeing this community grow.
It’s a real pleasure having you here.
And to everyone who’s reached out lately - thank you. Your messages, encouragement, and feedback mean more than you know.
Alright, let’s get into it ⬇️
Agents care about maths, not logos
Ask an agent to buy a 34‑inch monitor for the lowest cost. It will crunch shipping, FX, taxes - and increasingly, payment fees too, as they become more visible to agents evaluating cost‑to‑serve in real time. If a stablecoin bridge or Pay‑by‑Bank rail beats the card total, that is the path it will pick. No offence to the schemes - the agent just picks what’s cheapest and clearest.
What Visa and Mastercard actually shipped
An “agent” flag on network tokens so disputes don’t swamp support desks.
Spend caps you can set or pause in real time.
For Mastercard, a central list of approved bots. Visa leaves vetting to issuers.
Decent, but it reads like a virtual card with metadata.
Why smart contracts are the AI‑native fit
Smart contracts spell out every rule before a penny moves - amount, currency, delivery etc. They settle in seconds, expose the fee up front and remove the middlemen who add cost but no extra trust. Agents can simulate those contracts, confirm they meet the brief, then execute without asking anyone’s permission.
That determinism matches how software thinks. Cards were built for 1966’s America. Wrapping them in tokens helps, but it cannot change the fundamental flow of clearing and interchange.
Where cards still score (for now)
Global acceptance - roughly 150 million merchants already plugged in.
Built‑in consumer protection - chargebacks, zero‑liability, instant freeze.
Banks are already on the rails - they can market an “AI‑ready” feature without lifting a finger, and the incentives on scheme fees are familiar.
Ready to go - compliance, settlement and dispute journeys are battle‑tested.
The optimisation lens
AI tools are relentlessly optimising things. Agents are built to streamline and optimise.
Every extra basis point a rail shaves off cost, or every second it saves on settlement, shifts the probability that the agent picks it next time. In an open marketplace of payment APIs the leanest rail keeps winning until incumbents match the numbers or invent new value the code rails lack.
The AI age turns every purchase into a background API call. Intermediaries survive only if they add unique value. Visa and Mastercard own a superb risk engine and a global acceptance grid. Smart‑contract rails, meanwhile, deliver cost, speed and transparency that feel built for software.
Both camps will run in parallel for a while. The question is whether the existing network can bend its fee model enough before agents make the choice for us.
One more thing
For a deeper dive on why programmable money beats static rails, see my Money 2.0 piece.
About Dom Monhardt, founder of one-fs.com
I am a French technologist and product leader living in Dubai, with 15+ years of experience in building cutting-edge and innovative digital experiences.
I am interested in the intersection of business, design, and technology and am deeply passionate about the fintech and digital banking world.
Once agents stop caring about your branding and start caring about your bytecode, your moat is a memory.