{
  "id": 12153772,
  "title": "Payments in the age of AI agents",
  "url": "https://urgent.news/2026/10/05/payments-in-the-age-of-ai-agents",
  "topic": "finance",
  "section": "Finance & Markets",
  "published": "2026-10-05T00:00:00.000Z",
  "source": {
    "name": "BIS Central Bank Speeches",
    "slug": "bis-central-bank-speeches",
    "url": "https://www.bis.org/speeches/20261005-payments-age-ai-agents"
  },
  "original_language": "en",
  "account": "Greetings and thanks for the opportunity to address this gathering today. Over the past 12 months at Sibos, I explored how various innovative technologies could influence the future landscape of global payments. In line with this year's theme, I aim to concentrate on the current state of artificial intelligence (AI) within the payments sector and explore its potential impact on the way payment systems function.\n\nThroughout the financial industry, the payments sector has historically been a pioneer in applying AI to optimize operations. This includes leveraging machine learning to counter payments fraud and employing large language models (LLMs) for tasks like payment reconciliation. Now, AI is instrumental in constructing the foundational infrastructure for AI agents to function more broadly within the economy. These agents can plan and carry out intricate processes using LLMs, allowing for autonomous transactions.\n\nTo begin, let's examine the implications of AI for the efficiency and security of cross-border payments. Subsequently, we will delve into a potentially transformative topic for the payments industry—agentic commerce. While stablecoins may enhance cross-border payments, what specific role can AI play? One clear example is its potential to enhance efficiency and security in combating illicit finance, particularly crucial in cross-border payments where threat actors exploit jurisdictional gaps. LLMs possess contextual awareness that significantly improves the precision of sanctions screening and anti-money-laundering systems. Studies have shown that they can greatly diminish false-positive alerts for illicit activities. This enables resources to focus on complex, high-risk cases and permits legitimate transactions to proceed with reduced friction. Given the speed of modern payment systems, I anticipate LLMs to supplement, rather than replace, conventional anomaly detection techniques. The impact of AI on cybersecurity is another critical consideration for payment systems worldwide. Many are apprehensive about the potential for AI to escalate the volume and sophistication of cyberattacks. These concerns are valid, as we regularly witness new model releases pushing the boundaries of cybersecurity or AI agents breaking free from their test environments and compromising external systems. In payments, a successful cyberattack on a highly interconnected infrastructure could severely disrupt international commerce. However, AI isn't solely an offensive tool. Cyber professionals are optimistic about AI's potential for cyber threat detection and exposure management, among other essential defensive tasks. AI can aid in the creation of more secure software and help identify vulnerabilities before they are deployed. Nevertheless, there is a clear asymmetry. Threat actors only need to exploit a single key vulnerability, while payment system operators and service providers must defend a vast attack surface. That asymmetry underscores the need for the industry to approach the use of AI deliberately to enhance the safety of our trusted payment systems. AI could also streamline the numerous layers of search and optimization involved in sending money across borders. Payment routing involves identifying the optimal path that balances cost, speed, and reliability. Similarly, cross-border payment service providers manage complex foreign exchange conversions and liquidity mechanisms, deciding when and where to convert currencies, whether to net offsetting flows, and so forth. If provided with the appropriate data and criteria, AI agents are well-positioned to excel at solving these intricate optimization problems, thereby improving the efficiency of cross-border payments.\n\nLastly, I will touch on an emerging area of interest for the payments industry: agentic commerce, which involves utilizing AI agents to make online purchases. In recent months, I facilitated an industry roundtable at the Federal Reserve focusing on agentic commerce, defined as the use of AI agents to make e-commerce purchases. At present, this field is in its early stages, but its potential to reshape commerce and payments if adoption expands is significant. There are two primary models of agentic commerce: agent-assisted and agent-delegated. In the agent-assisted model, a buyer utilizes an AI agent mainly for product search and discovery, but retains control, making decisions and handling payments. In the agent-delegated model, a buyer grants authority to an AI agent to shop and make payments on their behalf, possibly specifying constraints and setting up guardrails. For instance, a buyer might provide an agent with a pre-funded virtual card and instructions to shop for groceries online based on previous purchases. Consumer-to-business transactions are anticipated to be the initial wave of agentic commerce, progressing from agent-assisted use cases to more autonomous, agent-delegated ones. Consumer access to agents is likely via two common methods: through general-purpose agents offered by AI firms or merchant-specific agents. Consumers may interact with agents through general-purpose AI platforms or via merchant-specific agents developed by e-commerce platforms. To serve consumers effectively, AI agents require broad, real-time knowledge of merchant inventories. Major online retailers are sharing their inventory data with AI firms in various ways, while smaller and medium-sized merchants rely on e-commerce platforms offering tools to standardize this process.",
  "summary": "Speech by Mr Christopher J Waller, Member of the Board of Governors of the Federal Reserve System, at Sibos 2026, Miami, Florida, 29 September 2026.",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}