AI in Finance and Insurance

Learning Objectives

• Identify the key applications of AI in financial services — fraud detection, algorithmic trading, robo-advisors, and AI-powered expense management
• Explain how insurance companies use AI for underwriting, claims automation, and computer vision for property damage
• Understand the regulatory and ethical dimensions of AI in financial decisions

Finance: Where AI Has Been Deployed Longest

Financial services — particularly algorithmic trading and fraud detection — were among the first industries to deploy machine learning at scale. The combination of structured data, high volume, clear success metrics, and enormous financial stakes made finance an early adopter.

Today, AI in finance spans from high-frequency trading to consumer banking to wealth management to expense management, touching virtually every aspect of how money moves.

AI in Banking and Financial Analysis

JPMorgan Chase: LLM Suite and AI at Scale

JPMorgan Chase employs more technologists than most technology companies and has made the largest AI investment of any bank. The company now reports 450+ AI use cases across the firm, with $18 billion committed to technology spending.

LLM Suite is JPMorgan's flagship AI product — launched in summer 2024, it now has 200,000 employees using it daily and won American Banker's 2025 Innovation of the Year Grand Prize. LLM Suite provides a secure internal environment using models from OpenAI and Anthropic for tasks across the bank — investment bankers create 5-page decks in 30 seconds, lawyers scan and generate contracts, and analysts process research at unprecedented speed.

IndexGPT is JPMorgan's AI system for investment theme selection — identifying investment opportunities across sectors and regions from natural language descriptions of economic trends. DocLLM is the bank's document understanding AI — trained specifically on the complex, structured-but-irregular format of financial documents: loan agreements, prospectuses, earnings reports, SEC filings. Financial documents are notoriously difficult for general-purpose LLMs because of their specialized terminology and complex table structures; DocLLM was designed to handle this specifically.

JPMorgan's AI scale: The bank reports AI benefits growing 30–40% annually. JPMorgan has stated that AI has saved billions of dollars annually in fraud prevention alone, and has begun deploying agentic AI for complex multi-step tasks across its operations.

Bloomberg Terminal AI

Bloomberg has integrated AI deeply into the Bloomberg Terminal — the platform used by virtually all professional investors and traders globally.

Bloomberg's AI features:

• Bloomberg Intelligence AI: Summarization and analysis of research reports across the Terminal's database of millions of documents
• Earnings analysis: Real-time summarization and analysis of earnings calls and filings as they happen
• Relationship mapping: AI-powered analysis of relationships between companies, executives, and events
• Natural language queries: Asking the Terminal natural language questions rather than requiring mastery of Bloomberg's proprietary query language

Bloomberg also developed BloombergGPT — a large language model trained specifically on financial data — and has integrated AI capabilities throughout its data and analytics products.

Intuit: AI in Tax and Accounting

Intuit (maker of TurboTax, QuickBooks, and Credit Karma) has deployed AI extensively throughout its consumer and small business financial products. In November 2025, Intuit signed a $100 million+ multi-year partnership with OpenAI — integrating TurboTax, QuickBooks, Credit Karma, and Mailchimp directly into the ChatGPT experience, allowing users to estimate tax refunds, find credit cards, and generate financial insights within ChatGPT.

TurboTax AI: Automated document extraction from W-2s, 1099s, and other tax documents; intelligent question answering throughout the tax filing process; detection of potential deductions based on financial data. Intuit reports that AI dramatically reduces completion time for tax returns.

QuickBooks AI: Intuit has rebranded its AI strategy around "agentic AI" — QuickBooks now deploys a "virtual team of AI agents" including Business Tax AI, Customer AI, and Accounting AI agents that autonomously handle accounting, finance, and sales tax tasks. Businesses report saving up to 12 hours per month. Over 3 million customers use Intuit's AI-powered tools.

Algorithmic Trading and Fraud Detection

Algorithmic trading using AI is not new — hedge funds and investment banks have been using machine learning for trading signals since the 2000s. What has changed is the sophistication of the models and the scale of deployment.

High-frequency trading (HFT) firms use AI for pattern detection in millisecond price movements. Palantir's AI analytics platforms are used by financial institutions for market analysis, risk modeling, and regulatory compliance.

Fraud detection is one of the most mature AI applications in finance. Every major card network (Visa, Mastercard) and bank uses AI to evaluate every transaction in real time — scoring the probability of fraud based on hundreds of features: transaction amount, location, merchant type, device fingerprint, behavioral patterns. Modern AI-powered fraud detection catches fraudulent transactions that rule-based systems would miss and produces dramatically fewer false positives that block legitimate transactions.

Robo-Advisors: AI-Powered Wealth Management

Betterment and Wealthfront are the leading robo-advisors — platforms that use AI algorithms to manage investment portfolios for retail investors at a fraction of the cost of traditional financial advisors.

The core AI capabilities:

• Automatic rebalancing: Continuously monitoring and rebalancing portfolios to maintain target asset allocations
• Tax-loss harvesting: Identifying opportunities to sell securities at a loss to offset taxable gains — a capability that requires constant monitoring impossible for human advisors to do manually
• Risk modeling: Dynamically adjusting portfolio composition based on market conditions and individual risk profiles

Betterment manages over $55 billion in assets (and acquired Ellevest's automated investing business in 2025); Wealthfront manages approximately $94 billion — having IPO'd on Nasdaq in December 2025 at $14/share, making it the first major robo-advisor to go public. Wealthfront's 1.4 million funded clients and $365 million in annual revenue demonstrate the scale that AI-powered wealth management has achieved. These platforms have democratized portfolio management techniques previously available only to wealthy clients of human financial advisors.

AI-Powered Expense Management

Brex and Ramp represent the new generation of AI-native corporate financial products. In January 2026, Capital One announced a $5.15 billion acquisition of Brex — the largest bank-fintech deal in history — signaling that traditional banks view AI-native financial tools as strategically essential.

Brex and Ramp use AI for:

• Automated receipt matching: AI matches every credit card transaction to the corresponding receipt, eliminating manual expense reports
• Anomaly detection: Flagging unusual spending patterns that may indicate fraud or policy violations
• Spend insights: AI-generated analysis of company spending patterns across vendors, categories, and departments
• Policy enforcement: Automatically flagging transactions that violate spending policies before they are processed

These companies have grown rapidly because the AI automation they offer directly addresses a painful, high-friction process (expense reporting) that every company with employees experiences.

Insurance: AI Across the Full Value Chain

Insurance is a data-intensive industry built on risk prediction — an ideal application domain for AI. AI is transforming insurance across underwriting, pricing, claims processing, and fraud detection.

Lemonade: AI-First Insurance

Lemonade is the highest-profile AI-native insurance company. Founded in 2015, it built its entire insurance platform on AI from the ground up — using AI for underwriting, fraud detection, and claims processing. In FY2025, Lemonade reached $738 million in revenue (up 40% year-over-year), nearly 3 million customers, and $1.24 billion in in-force premium — with a path to EBITDA profitability targeted for Q4 2026.

Lemonade's AI claims bot, Jim, can process and pay certain claims in seconds — a fraction of the days or weeks typical in traditional insurance. For straightforward claims (a stolen bicycle, water damage from a burst pipe), the AI reviews the claim, checks against the policy, and initiates payment without human intervention. Lemonade's loss adjustment expense (LAE) ratio dropped to 7% in 2025 — a direct result of AI automation reducing the cost of processing each claim, even as claim volumes grew.

Lemonade's fraud detection AI cross-references claims against hundreds of behavioral signals — including the emotional state of the claimant during video submission — to identify potentially fraudulent claims for human review. The company has expanded beyond renters and homeowners insurance into pet, auto, and European markets.

Root Insurance: Telematics and Behavioral Pricing

Root Insurance uses AI to price auto insurance based on actual driving behavior rather than demographic proxies. The Root app measures acceleration, braking, cornering, phone use while driving, and time-of-day patterns — generating an AI-computed risk score that determines the customer's premium. Root has expanded its data sources beyond phone sensors with connected vehicle data partnerships — Toyota and Lexus vehicles now feed driving data directly to Root via OEM integration.

Root reached $1.52 billion in revenue in 2025 (up 29%) and posted its first annual profit — a significant milestone for a company that was previously burning cash. The company now operates in 36 states.

This behavioral approach is more accurate than traditional actuarial methods (which rely on demographic factors like age, gender, and location that are imperfect proxies for driving risk) and is argued to be more equitable (basing price on actual behavior, not demographic characteristics).

Tractable: Computer Vision for Claims

Tractable applies computer vision AI to damage assessment — the process of evaluating physical damage to vehicles and property after accidents or disasters.

Traditional claims adjustment requires a human adjuster to physically inspect damaged vehicles or property, estimate repair costs, and authorize payments. Tractable's AI can analyze photographs of damaged vehicles to produce accurate repair estimates in seconds — the same analysis that takes an experienced human adjuster hours or days.

The result: Claims processing times measured in hours rather than days or weeks. Major insurers including Tokio Marine, IAG, and Ageas use Tractable for AI-assisted claims processing.

Cape Analytics: AI for Property Risk

Cape Analytics (acquired by Moody's Corporation in January 2025) uses satellite and aerial imagery with AI to assess property risk without physical inspection. Now part of Moody's Intelligent Risk Platform, Cape Analytics serves nearly half of the top 50 US property insurers. For insurance underwriters, understanding the condition of a property — roof age, solar panel presence, proximity to trees, pool presence — is essential for accurate pricing.

Traditional property inspection requires scheduling and sending an adjuster to the physical location. Cape Analytics uses AI to extract this information from aerial imagery at a fraction of the cost and in minutes rather than days. Moody's acquisition validates the technology and integrates it with one of the world's largest financial data companies.

Key Takeaways

• Finance was an early and deep adopter of AI — fraud detection and algorithmic trading have used ML for decades; LLMs are now being deployed for research, analysis, and document understanding
• JPMorgan (LLM Suite with 200K daily users, IndexGPT, DocLLM), Bloomberg Terminal AI, and Intuit ($100 million+ OpenAI partnership, agentic AI agents) represent the depth of AI integration across financial services
• Robo-advisors (Betterment ~$55 billion, Wealthfront ~$94 billion with 2025 IPO) have democratized AI-powered portfolio management; AI expense management (Brex — being acquired by Capital One for $5.15 billion, Ramp) has automated a high-friction business process
• Insurance AI spans the full value chain: Lemonade ($738 million revenue, AI-driven 7% LAE ratio), Root ($1.52 billion revenue, first profit, Toyota/Lexus data), Tractable (computer vision claims), Cape Analytics (acquired by Moody's)
• AI in financial decisions faces legitimate regulatory scrutiny around fairness, transparency, and potential for discriminatory outcomes — explainability and audit trails are required