Qonto builds Claude-powered agents that run payroll 5x faster and cut invoice creation time 3x

Overview

Qonto, a European business banking platform serving over 600,000 businesses across eight markets, built AI agents on Claude Opus 4.5/Sonnet/Haiku via Amazon Bedrock that prepare bulk payroll transfers, draft client invoices from a CSV, and answer cash-flow questions, with the customer always giving final approval before money moves. Since launching six weeks after the November 2025 project start, the agents send bank transfers 2x faster, cut invoice creation time 3x, run payroll 5x quicker, and have processed customer-approved transfers as large as €100,000.

The challenge

Qonto customers spend on average up to 8 hours every month on financial admin tasks: manually running payroll transfers one pay slip at a time and re-keying client invoices one at a time from a CRM or spreadsheet, work that became a bigger pain as transaction volume grew, while cash flow lived in hand-built spreadsheets with account data going unused for decisions.

The solution

Qonto's 15-person AI Lab built agents into the app, routing each step by complexity to the right model on latency, performance and cost: complex missions run on Claude Opus 4.5 and Sonnet while simpler steps go to smaller models including Haiku, accessed through Amazon Bedrock with GDPR and EU AI Act compliance handled at the provider level, with an operator agent handling high-frequency tasks like payroll and invoicing end to end and an analyst agent answering questions about past and present transactions.

Reported business value

Bank transfers now send 2x faster with no manual beneficiary search, amount entry or reference copying; client invoice creation takes 3x less time; payroll runs 5x quicker, turning a 15-to-20-slip run into one bulk transfer; customers have delegated transfers as large as €100,000; and one customer generated over 500 monthly invoices in a single batch.