Payouts.com co-founders say the future of agent payments combines stablecoin rails with programmable control layers built for enterprise trust.
Summary
- Payouts.com CEO Leor Ceder says programmability, not wallets alone, will define which AI agents enterprises can trust by 2027.
- Co-founder Barak Hirchson lists five non-negotiable controls that make autonomous agent spending safe and auditable at scale.
- Stablecoins win in cross-border and machine-to-API micropayments; programmable infrastructure determines which rail gets used everywhere else.
Payouts.com co-founders Leor Ceder and Barak Hirchson say the next wave of AI agent commerce runs on stablecoin rails, and on the programmable control layer built on top of them. In their view, wallets are a necessary foundation, but the durable enterprise value sits in what governs them.
The position adds a critical dimension to the wallet-led narrative dominating agent payments today. Juniper Research forecasts cross-border B2B stablecoin payments will hit $5 trillion by 2035, up from $13.4 billion in 2026, with B2B taking 85% of total stablecoin transaction value.
Where stablecoins win and where smart rail selection matters
Hirchson, Payouts.com’s chief solutions officer, said rail selection is decided by the recipient: country, payment method, urgency, amount, and cost all factor in. Stablecoins win cleanly in two scenarios.
The first is cross-border versus SWIFT, where wire fees and FX spreads can eat 4 to 5% of a transaction. The second is machine-to-API micropayments, where the x402 standard already routes pay-per-call API invoices in stablecoin. Crypto.news reported that AI agents have settled $73 million across 176 million transactions on crypto rails, with USDC handling 98.6%.
“PIX clears in under ten seconds in Brazil for free, UPI handles hundreds of millions of transactions a day in India at near-zero cost,” Hirchson said. “The agents that scale are the ones that can pick the right rail per transaction, not the ones locked into a single rail based on what their limited wallet supports.”
The five non-negotiable agent controls
Hirchson laid out five controls he said are non-negotiable before companies let agents transact autonomously: scoped credentials, hard spend caps enforced at the protocol level, cryptographically signed mandates, idempotency at the payment layer, and a fail-closed posture.
“This is what programmable spending actually means. You define the envelope once, the infrastructure enforces it forever, and the agent operates freely inside it,” he said. “Is the industry building these fast enough? Not uniformly.”
Some wallets shipped recently include hard caps and signed mandates, he said. Others ship with an API key and a balance, which he called the worst-case configuration for a compromised key.
What the agent payment stack looks like by 2027
Ceder said the interesting question by May 2027 will not be which stablecoin wins. It will be programmability: how granularly enterprises can define what an agent is allowed to do, how reliably that policy is enforced, and how cleanly compliance can be proven after the fact.
“The wallet wars happening right now will look the way the browser wars look in retrospect: necessary, formative, and not where the durable value got captured,” Ceder said. The compliance layer must be built into the infrastructure rather than the agent, with every payment passing a cascade of principal, account and jurisdiction checks before any money moves.
Coinbase and Cloudflare have built the x402 protocol into a fast-growing settlement rail for agents, with the standard recently joining the Linux Foundation. AWS embedded x402 into Amazon Bedrock AgentCore Payments earlier this month, while Solana and Google launched Pay.sh as a parallel route.
For Payouts.com, the bet is that the control layer above those rails is where enterprise spend will land. The agent stays autonomous. The envelope around it does not move.
