Monzo engineers demo five trust principles for financial AI agents
AI Engineer conference published a video where Monzo engineers Francesco Galletta and Felippe Felisola Caso build a financial analyst agent for wedding budgeting. They present five design principles for trustworthy AI: shared domain, consistency, evidence, accountability, and verification, explaining why calculations belong in code.
ImportanciaLocalEvidenciaE2 no replicadoAnálisisRápido
The AI Engineer conference released a video titled "Stop Chatting With Spreadsheets" on October 17, 2026, featuring Monzo engineers Francesco Galletta and Felippe Felisola Caso. The resource addresses the core challenge in financial AI applications: while fast answers are easy to obtain, answers that can be trusted with money are difficult.
Using a real-world problem—paying for a wedding—as a case study, the speakers build a financial-analyst agent harness live. According to the episode description, they outline five design principles for trustworthy AI: shared domain, consistency, evidence, accountability, and verification. A key point highlighted is that calculations should reside in code rather than within the model's internal reasoning.
For readers interested in AI implementation in finance, this video offers a reference path from concept to execution. Although a full transcript is not provided, timestamps indicate detailed discussions ranging from why linear chats fail to the application of specific principles. This resource serves as background material for professionals seeking to understand how structured design enhances the reliability of AI-driven financial decisions.