AI, Data and Automation in Finance

Finance was a data business long before "AI" was a marketing word. A bank is a machine for deciding whom to trust with money; a trading firm is a machine for turning information into prices; an insurer is a machine for pricing uncertainty. Statistical credit scoring dates from the 1950s (Fair, Isaac and Company was founded in 1956), algorithmic trading from the 1980s, and neural-network card-fraud detection from the early 1990s. What changed in the 2010s was the scale of data and computing, and in 2023–26 the arrival of large language models (LLMs) that can read, write and reason over the unstructured text in which most financial work is done. The economics follow a simple rule: AI creates value in finance where decisions are frequent, outcomes are measurable and feedback is fast (fraud scores, trading signals, collections), and creates risk where outcomes are slow, rare or politically sensitive (credit through a recession, fair lending, advice). This section covers quantitative trading, credit models and alternative data, fraud and financial-crime detection, generative AI and LLMs in banks, operations automation, model risk management (including the April 2026 replacement of the Fed's SR 11-7 guidance), the regulation of AI in finance in the US (with EU and UK comparisons), data as a business, and scoring thin-file Americans with cash-flow data. It closes with cases, pitfalls and how experienced practitioners judge AI claims.