UK digital bank Starling has deployed an artificial intelligence agent that investigates transaction monitoring alerts end to end in 17 seconds. The agent produces a decision together with a full audit trail. It is built directly into Starling's banking system, so it sees each transaction's history and context in real time without external integrations. It can close an alert, escalate it to a person or file a suspicious activity report with the National Crime Agency.

What it means: Anti-money laundering checks are among the most expensive parts of running a bank, because monitoring systems raise many alerts and most turn out to be innocent. Each one still needs an analyst to review it, which costs time and causes delays for customers. Starling shows that an agent can do most of the preliminary work almost instantly, as long as it has full access to the data and leaves a clear trail of how it reached its view. The crucial question is accountability: when an agent can close alerts or file reports with the authorities, the bank must be able to explain every decision to its supervisor. For Greek banks and payment institutions, facing growing demands from the Bank of Greece and the new European authority AMLA, such tools could lower compliance costs. Adopting them, however, needs good data quality, documented models and human review for the most serious cases. The example also shows that integration into the core system makes the difference in speed.

Read the original on Finextra

Related service: myDATA integrations

Related stories