The theme of Sibos this year is digital finance for AI-driven economies, and the most concrete version of that phrase is an AI agent that pays. A consumer’s assistant that books the trip and settles the invoice. A corporate agent that manages supplier payments against a treasury policy. The card networks have published frameworks for authenticating agent-initiated transactions, and NPCI in India released an agentic payments framework in September. The direction is set.

Most of the commentary is about the agent. We think the harder problem sits underneath it.

An agent does not care which rail it uses. It will pay by card if the merchant takes cards, by an instant account-to-account rail if that is cheaper or faster, by wire if the amount demands it. The institution behind the agent’s funding account has to route across all of those, apply the customer’s limits and the bank’s policy, authenticate a request that carries no human session behind it, and score it for fraud, in real time, every time. Today those functions live in four or five systems that were never designed to make one decision together.

That is an orchestration problem, and it is why payment orchestration stops being a plumbing topic and becomes the control layer of agentic commerce. The orchestration layer is the one place that sees the instruction before it hits a rail and knows enough about the customer, the counterparty and the policy to say yes, no, or ask. Fraud scoring belongs inside it rather than beside it, because a compromised or manipulated agent will produce transactions that look legitimate one at a time and only reveal themselves in the pattern.

Rather than a product pitch, here is the checklist we would put in front of a bank’s payments and risk leadership before the traffic arrives.

  • Can one policy be expressed once and applied across cards, instant rails and wires, or is it maintained separately in each?
  • When a transaction arrives with an agent credential instead of a customer session, which system authenticates it, and against what?
  • Can routing take fraud risk into account, so that a high-risk instruction can be moved to a rail with recall, held, or stepped up?
  • Is every agent-initiated payment scored against the customer’s behavioural baseline, or only on the transaction’s own attributes?
  • Does the audit record show who authorised the agent, what limits applied and why the decision was made, in a form a dispute team can use?
  • How long does the whole path take, and does it fit inside the settlement window of the fastest rail you support?

Most institutions will answer “in parts” to most of these. That is not a criticism. The stack was built for humans pressing buttons. The point is that the gap is an orchestration gap, and closing it does not require replacing the core or the fraud platform. It needs a layer above both that can connect the rails, orchestrate the decision and apply control, added to what is already running.

We built that kind of layer for national-scale real-time payments, with the fraud scoring running inside it. Agentic payments are a new source of traffic for the same architecture, not a new architecture.

About RS Software
As the payments ecosystem transitions toward an AI-driven, real-time, and increasingly autonomous model, where TRUST is the new currency, RS Software (India) Limited's product portfolio modernizes the most critical areas across the value chain, leveraging decades of work building mission-critical payment infrastructure globally. Each product is purpose-built to address specific structural shifts in the market while collectively enabling a unified, AI-led financial ecosystem. RS IntelliEdge™ is designed to address evolving transaction-level controls and shift to continuous, identity-centric intelligence systems, protecting approximately 95% of India’s digital payment transactions and screening billions of transactions each month. RS DigitalEdge™ is designed to enable real-time routing, optimization, and decisioning across payment flows. Payabbhi® is designed to enhance the acceptance and acquiring layer, transforming it into a data-driven revenue engine. www.rssoftware.com