As money becomes programmable, the decisions behind it are becoming autonomous. AI agents are starting to optimize liquidity, move collateral, price credit and execute treasury operations in real time, working alongside, and increasingly ahead of, human teams. For banks, the competitive edge is shifting from processing transactions to controlling the intelligence that decides how and when they happen.

This session brings together TradFi executives and leaders from AI labs to explore what it takes to put AI agents to work on the balance sheet. They’ll cover agentic treasury and liquidity management, programmable lending and automated credit decisions, and collateral that moves on its own based on predefined rules. They’ll also address the controls that make this safe—including authority limits, model risk management, explainability, and audit trails for autonomous actions. Panelists will also consider how far autonomy should extend, what still requires human judgment, and which institutions will own the intelligence layer of banking. Specific discussion will also include:

• Where are AI agents making balance-sheet decisions today, and where are banks drawing the line?
• How do you set, enforce and audit the authority of an autonomous agent?
• Will banks build this intelligence themselves, partner with AI labs, or risk ceding it?

Institutions won’t scale on-chain finance until they’re confident it will stand up to supervisory scrutiny. As programs move from pilot to production across banking, payments, capital markets and asset management, compliance is shifting from after-the-fact review to controls built directly into the infrastructure. Identity becomes portable, monitoring becomes continuous, and rules execute in code alongside the transactions they govern.

This session will examine what “examiner-ready” means in practice. Panels will cover embedded KYC/AML, on-chain identity and allowlisting, real-time transaction monitoring, smart contract audits and governance, and how institutions document controls that execute automatically. Looking ahead, they will also consider how supervisory expectations are evolving as digital-asset frameworks take shape, where regulatory gaps remain, and how programmable compliance could lower costs while raising the bar for trust across the financial system. Other notable areas of the conversation:

• What do supervisors expect to see when they examine a production on-chain program?
• How do you prove that controls embedded in code are working as intended?
• Can programmable compliance reduce cost and risk at the same time, and what still requires human judgment?

The next customer to initiate a payment may not be a person. AI agents are beginning to shop, procure, pay invoices and transact with other machines on behalf of consumers and businesses. Today’s payment systems were built around human intent, authentication and disputes, and that assumption is starting to fail.

This panel discussion will examine the leaders that are building the rules of agentic commerce, looking at how agents are authenticated and authorized, how spending mandates and limits are set and enforced, and who bears liability when an agent makes a mistake or is manipulated. 

Panelists will also discuss how disputes and chargebacks work without a human at the point of sale, and which agent payment protocols are emerging across card, real-time and stablecoin rails. Looking ahead, they will consider what institutions must have in place before agent-initiated payments reach scale, and who will own the trust relationship between customers, agents and the payment system. The conversation will also tackle:

• How do you verify that an agent is acting within its mandate, and prove it after the fact
• When an agent pays wrongly, who is liable: the customer, the agent provider, the merchant or the bank?
• Which rails and protocols are best suited to machine-to-machine payments at scale?

The future of payments is many rails running side by side. Cards, ACH, FedNow, RTP, stablecoins and tokenized deposits increasingly coexist within the same institution, and customers expect instant, low-cost, transparent payments no matter what moves underneath. The competitive edge now lies in orchestration: deciding which rail each payment should take, and making the whole system work as one.

This session brings together leaders running multi-rail operations in production to dive into intelligent routing based on speed, cost, compliance and liquidity; reconciliation across rails that settle on different timelines; real-time treasury visibility; and how pricing and economics change when rails compete for the same transaction. Panelist will also look ahead to AI-driven routing engines that make payment rails effectively invisible, and to who will own the orchestration layer as payments become embedded infrastructure for commerce, treasury and capital markets. Other discussion points will include:

• How to decide which rail a payment takes, and who or what makes that call?
• What breaks in reconciliation and liquidity management when rails settle on different clocks?
• As rails commoditize, where does payments revenue come from next?