AI in Financial Services: From Trading to Underwriting

Introduction

Artificial intelligence is transforming financial services more rapidly and comprehensively than any technology since electronic trading. Global AI spending in banking is projected to exceed $110 billion by 2026, with 75% of banks holding more than $100 billion in assets expected to fully integrate AI strategies by 2025. The impact spans every FIG sub-sector: algorithmic trading at quantitative firms, credit underwriting at consumer lenders, fraud detection at commercial banks, automated claims processing at insurers, and portfolio construction at asset managers. For FIG bankers, AI creates dual relevance: it is both a coverage theme driving fintech valuations and M&A activity, and an operational force reshaping how every financial institution they advise competes, underwrites, and serves customers.

The most visible AI transformation in financial services is the enterprise-wide rollout of generative AI tools at major banks. JPMorgan Chase deployed its LLM Suite platform to over 200,000 employees, with 125,000 daily active users leveraging the tool for research synthesis, document analysis, and client communication drafting. Goldman Sachs launched the GS AI Assistant firmwide in mid-2025 (after piloting with 10,000 employees), reporting that it cut deck preparation time by 50%, translating to thousands of reclaimed analyst and associate hours. Morgan Stanley's GPT-4-powered "AI @ Morgan Stanley Assistant," which draws on approximately 100,000 research reports and documents, reached 98% adoption among wealth management advisor teams within months of its September 2023 launch.

Generative AI in Banking

Generative AI refers to large language models (LLMs) and related systems that can produce new text, code, analysis, and structured outputs rather than simply classifying or scoring existing data. In banking, generative AI applications include drafting client memoranda from raw financial data, synthesizing hundreds of pages of regulatory filings into executive summaries, generating investment theses from earnings transcripts, and automating portions of due diligence document review. The distinction from traditional machine learning is that generative AI handles unstructured tasks (writing, summarizing, reasoning) while traditional ML excels at structured prediction tasks (credit scoring, fraud detection, pricing). Both are being deployed simultaneously across financial institutions, but generative AI has captured outsized attention because it directly augments the work of bankers, analysts, and advisors.

These deployments are not experimental. They represent a structural shift in how financial institutions operate. JPMorgan's IndexGPT uses OpenAI's GPT-4 to create thematic investment baskets by scanning news, earnings calls, and market data for relevant companies. The productivity gains are measurable: Goldman estimates its AI tools will save the equivalent of hundreds of full-time roles in back-office and middle-office functions over a multi-year horizon. The competitive dynamic is self-reinforcing: banks that deploy AI effectively attract better talent (analysts prefer firms where AI handles routine work), serve clients faster, and generate higher revenue per employee.

AI-driven credit underwriting represents the most direct intersection of technology and FIG deal flow. Traditional credit scoring relies on FICO scores and limited historical data. AI models trained on alternative data (payment patterns, cash flow trends, extended credit histories, transaction behavior) can evaluate creditworthiness with significantly greater precision.

Zest AI, named one of CNBC's top fintech companies in 2025, reports that banks and credit unions using its platform approve 25% more loans at lower interest rates without increasing default risk. Upstart, which received a CFPB no-action letter in 2017 (a watershed moment for AI in lending), demonstrated that machine learning models could outperform FICO-only approaches in predicting default probability, though the CFPB terminated the no-action letter in 2022 when Upstart requested significant model changes without regulatory review.

Algorithmic and AI-driven trading have moved from niche quantitative strategies to dominant market forces. Citadel Securities uses AI to enhance high-frequency trading execution, analyzing market data in real time across thousands of simultaneous trades. Two Sigma manages over $58 billion in assets using machine learning, distributed computing, and predictive analytics. Renaissance Technologies' Medallion Fund remains the gold standard of AI-driven systematic trading.

The competitive landscape has shifted: 68% of hedge funds now employ AI for market analysis and trading strategies, and 65% of global financial institutions use machine learning for portfolio management. The distinction between "quantitative" and "fundamental" investing is blurring as traditional asset managers layer AI tools onto fundamental research processes, using NLP to parse earnings calls, satellite imagery to track retail foot traffic, and alternative data to generate alpha signals that complement analyst judgment.

Fraud detection is the most mature and universally adopted AI application in financial services. Over 87% of global financial institutions have deployed AI-powered fraud detection systems, up from 72% in early 2024. The performance improvement over legacy rule-based systems is stark: AI detects 70-90% more suspicious activity while reducing false positives by 80-90%. Legacy anti-money laundering (AML) systems generate 90-95% false positive alerts, creating enormous compliance costs and operational drag that AI directly addresses.

HSBC monitors 900 million transactions monthly across 40 million accounts using Google Cloud's AML AI platform, combining machine learning with transaction surveillance at a scale impossible for human analysts. The Financial Action Task Force (FATF) has endorsed advanced analytics for AML, reflecting regulatory consensus that AI, when properly governed, enhances compliance program effectiveness.

Insurance is being reshaped by AI across underwriting, claims, and pricing. Progressive, with over 35 million policies and two decades of proprietary driving data, uses hybrid models blending generalized linear models with neural networks for telematics analysis, achieving 9% more accurate risk pricing that enables lower premiums for the safest drivers. Lemonade's AI Jim claims bot handles approximately one-third of claims autonomously, making payout decisions in seconds and contributing to a gross loss ratio improvement of 12 points year-over-year (to 67% in Q2 2025). Root Insurance prices auto insurance entirely through telematics-based AI underwriting.

Robo-advisory platforms have crossed $1 trillion in global AUM, with Vanguard Digital Advisor leading at approximately $312 billion, followed by Empower at $200 billion, Schwab Intelligent Portfolios at approximately $81 billion, and Wealthfront at $75 billion. Wealthfront achieved profitability through its low-cost, all-digital model (maintaining a 0.25% advisory fee), validating the scalability of AI-driven wealth management. The hybrid model (combining automated portfolio management with human advisor access) is emerging as the dominant approach, particularly for the mass affluent segment targeted by bank wealth management divisions and RIA platforms.

The RegTech market, projected to reach $82 billion by 2033 (from approximately $13 billion in 2023), applies AI to the regulatory compliance burden that is unique to financial services. Natural language processing (NLP) scans regulatory texts, cross-references them with institutional operations, and flags required changes. KYC automation uses real-time data feeds and behavioral analytics to maintain continuous compliance rather than periodic reviews. Agentic AI systems capable of executing multi-step compliance workflows autonomously are projected to reshape AML and KYC operations throughout 2026.

AI is both driving and reshaping FIG M&A. AI-focused fintech companies command premium valuations: AI startup Series A rounds average $51.9 million (30% higher than non-AI counterparts), and AI company revenue multiples average 23.4x versus 4.7x for broader fintech. The vertical AI market (industry-specific AI solutions) was valued at $10.2 billion in 2024 and is projected to reach $115 billion by 2034 at a 24.5% CAGR.

Notable acquisitions reflect the trend: Worldpay acquired Ravelin (AI-driven fraud prevention), Incode acquired AuthenticID (AI-powered identity verification), and BlackRock's $3.2 billion acquisition of Preqin was driven partly by the data infrastructure necessary to apply AI across private markets. The SEC's 2025 examination priorities explicitly include AI integration in portfolio management, trading, marketing, and compliance, signaling that regulatory scrutiny of AI in financial services will intensify alongside adoption.