banking technology teams

Abstract: This paper presents Making Tax Smart with Open Banking (MTS-OB), a novel DLT-based architecture designed to integrate real-time VAT compliance into the existing payment system through split payments and Open Banking. Motivated by the substantial VAT gap, the research investigates how leveraging the informational value embedded within payment transactions can reduce the tax administration burden. MTS-OB, a proof-of-concept application built on R3 Corda, facilitates automated, real-time VAT split payments (VSP) without disrupting current payment infrastructure. MTS-OB integrates with existing banking systems via Open Banking APIs, enabling secure information management for VAT collection, reporting, analytics, and automated compliance enforcement. The system uses traditional bank deposit money as a real-time information sensor thereby extending its functionality to include "money as a medium of real-time taxation". Scalability testing demonstrates the system's ability to maintain operational stability under increasing transaction loads, illustrating its potential for real-world deployment. The key finding is the necessity of redefining money to include its informational value for real-time tax compliance, requiring a redesign of the money system while adhering to existing regulations and standards. This research demonstrates the feasibility of integrating tax functionality directly into payments, offering a path towards "hard to get tax wrong" compliance and a potential reduction in the VAT gap without compromising financial stability.

Abstract: Corporate and investment banking teams use presentations to support credit decisions and advise clients on financing and transactions. Producing these decks requires reconciling financial data, tracing sources and turning analysis into a recommendation. We retrospectively study the development of an agentic harness combining a 27B language model, financial calculations, narrative templates and validation checks. LLM judges guide engineering changes and assess the resulting decks, raising the question of whether higher scores reflect better documents or changes in grading. In shared-session text-only grading with template markers removed, five judges score the complete system 20.4 to 33.6 points out of 95 above the same model generating directly from a short prompt. Every judge scores the system higher on all seventeen development deliverables. Margins against direct Opus generation from a short prompt range from -4.7 to +0.8 points. Judges agree on broad progress across development rounds but agree less on final-deck rankings than on pooled scores. Repeated grading also shifts scores on unchanged decks, making small improvements difficult to distinguish from judge variability.