Satispay, Italy's leading independent payment network with 6 million consumers and more than 450,000 merchants, now generates over 75% of the code committed to its repositories each month with Claude, and more than 90% of its engineers use Claude Code. The team completed the Java 8 to Java 21 and major Spring upgrade of its core transaction service, the system that runs every payment on the network, in under four days against a four-week estimate. Data transformation lambdas that took three to five days now take under an hour, and goals set for an 18-month code update roadmap were completed in 7 months, fast enough that an external consulting engagement brought in for the work was no longer needed.
"Claude Code's output, both the code it wrote and the reviews it produced, was considered best across the board," said Chief Technology Officer Fabio Rapposelli.
When Rapposelli joined in 2025, the engineering team was weighted toward earlier-career engineers and the codebase had some legacy services that were hard to maintain. Early-career engineers spent more time reading than writing, and senior review had become the bottleneck, carrying teaching, quality gate and context transfer at once.
In mid-2025 Satispay ran a structured 30-day evaluation comparing Claude Code against another AI coding tool on real tickets across its Java and Spring services. The rubric covered code generation quality, refactoring, adherence to internal patterns, multi-file context understanding, test generation and documentation, and Claude won on every dimension except IDE integration. Rollout took 30 days from decision to full coverage and was handled by IT support. Claude Code ships to every engineering laptop through Satispay's managed device fleet, so it is available on an engineer's first day. Engineers use it to write new code, modify existing services, navigate unfamiliar parts of the codebase and run a Claude review pass before human review.
The team packages reusable assets: standard Java scaffolding for new services and tests starts from an encoded pattern, and modernization approaches are packaged into subagents so the next team inherits the work. AWS Lambda functions that engineers used to write by hand are now generated from a specification by an internal subagent connected to an MCP server. Satispay added AI proficiency to its engineering performance framework, and engineers own what lands in the repository. Claude Cowork has also spread across several business units, and requests for Claude skills came in at roughly 20 a day from finance, legal, marketing and operations.
Satispay is now planning to ship roughly 50% more features in the first half of the year. Two pilots are running: issue-to-PR automation, in which Claude drafts pull requests from tickets filed by other teams, and agentic fraud investigation, with people kept in the loop on the decisions that warrant it.