Banks and exchanges adopt AI to monitor market infrastructure

A five-year, $4.8 million contract with one of the world's largest investment banks is pushing artificial intelligence deeper into the infrastructure that underpins trading. The deal, one of three such agreements signed recently, tasks an AI system with monitoring trade-routing, settlement timing and order-processing latency across the bank's internal systems. The technology sits between the front office and back office, watching the pipes that connect traders to exchanges and clearinghouses.

Over the past two years, financial firms have focused AI on front-office trading strategies and back-office compliance. This contract turns the tool on the infrastructure itself. The system is designed to flag anomalies in message flows, detect settlement delays before they compound and identify routing errors that could lead to failed trades. The bank declined to comment on the specifics of the deployment, according to a person familiar with the contract.

The push extends beyond banks. Several exchange operators have also started embedding AI into their own infrastructure monitoring. Nasdaq, for instance, already runs machine-learning models to detect market manipulation and is now testing similar models for latency monitoring and capacity management. The concept is the same: catch infrastructure problems before they cascade into a trading halt or a settlement failure.

For the vendors that supply market data and surveillance software, the trend opens a new revenue stream. Companies such as Nasdaq and Intercontinental Exchange, which operate both exchanges and technology platforms, are well positioned to sell AI monitoring tools to their banking clients. The $4.8 million contract runs through 2029 and covers systems integration, training and ongoing maintenance.

The adoption raises practical questions for traders and risk managers. If an AI system identifies a routing error at 9:59 a.m., who gets the alert and how fast can they act? The bank has set up a dedicated response team tied to the monitoring platform, the person said. The result, in theory, is fewer settlement fails and lower operational risk across the trading day.

The other two contracts, each smaller than the first, involve a European exchange and a custody bank. Terms were not disclosed. Together, the three agreements suggest that infrastructure monitoring has become a real, budgeted use case for AI in finance, not just a pilot or a proof of concept.