Snowflake and Amazon Quick cut AML alert review from an hour to five minutes

Amazon and Snowflake have teamed up to speed up one of the most labor-intensive processes in financial services: reviewing suspicious transactions to detect money laundering (AML). The result beat expectations: analysis time fell from an hour and a half to under five minutes.

Why this was a problem

Banks and financial companies receive thousands of alerts every day about potentially suspicious transactions. Each alert must be checked manually: an analyst looks at payment history, customer data, geography and spending patterns. That takes from half an hour to an hour and a half per case.

The process is expensive, slow and prone to human error. Compliance teams at large banks sometimes sit in a queue waiting for reviews. Transactions get delayed, customers complain and costs rise.

Most importantly, it is impossible to check every alert with the same thoroughness. Teams have to prioritize, which means something can slip through.

How Amazon Quick + Snowflake Cortex works

Amazon Quick Flows and Snowflake Cortex AI connect through the Model Context Protocol (MCP). The workflow automatically collects information about the customer and the transaction, passes it to the AI and receives a recommendation on the alert's status. The analyst only has to confirm or reject the machine's decision.

The process looks like this: the system pulls data from payment systems, the customer's history, their KYC (Know Your Customer) profile and the geography of the transactions. Snowflake Cortex analyzes all of it in context and returns a risk rating: high, medium or low. When needed, the system suggests additional steps or, if the risk is clearly low, closes the alert automatically.

What the system does

• Collects data from multiple sources (payment systems, transaction history, KYC databases)
• Analyzes customer behavior through Snowflake Cortex AI
• Ranks risk automatically, with an explanation
• Prepares a report for the analyst
• Logs everything for compliance and audit

All of this happens in a few seconds instead of an hour to an hour and a half.

Numbers from testing

Amazon and Snowflake ran tests and reported these results:

• Review time per alert: from 30–90 minutes to under 5 minutes
• Analyst workload: reduced by 80–90%
• Throughput: one team can process 10 times more alerts
• Accuracy: the AI catches patterns that humans miss

Results depend on alert complexity, data volume and workflow configuration, but even in the worst case the gain is substantial.

What this means for financial services

Financial institutions get a tool to detect fraud and money laundering faster while freeing people from routine work. This matters most when volumes are growing and compliance specialists are hard to find.

For AWS and Snowflake, this is another example of how AI can make enterprise processes more useful, and of MCP becoming a standard for integrating different systems.