Experian's tech chief defends credit scores: 'We're not Palantir'

What It Covers

Alex Lintner, Experian's CEO of Technology and Software Solutions, explains how the credit reporting company manages data on 247 million Americans while implementing AI systems. He defends credit scoring as essential for economic prosperity, details security practices including double encryption and data sharding, and addresses consumer trust concerns while deploying 200 AI agents across products.

Key Questions Answered

• •AI Governance Framework: Experian uses AI primarily for governance, explainability, and human oversight rather than direct decision-making. Small language models detect model drift in lending products by comparing predicted versus actual loan losses, then alert data scientists to adjust variables. All AI outputs require human validation by data scientists before production deployment to prevent hallucinations and bias.
• •Security Architecture at Scale: Experian implements data sharding across 25 separate encrypted locations, meaning attackers must break 25 encryption keys and reassemble fragments to access one person's complete profile. Account numbers receive double encryption and never leave internal systems except for matching purposes. The company operates as the only real-time credit bureau globally, enabling instant score updates versus 30-day delays at competitors.
• •Credit Access Innovation: Experian Boost allows consumers to add recurring payments like utilities, streaming services, and cell phone bills to credit scores for free, benefiting both consumers and lenders. This addresses the fundamental problem that immigrants and underbanked populations face: no lending history means no credit access, creating economic barriers. Traditional credit bureaus only count lending history, excluding millions from credit qualification.
• •Federated Organization Model: Experian employs 23,000 people with 11,000 in technology roles. The structure separates central functions (technology, security, finance) that maintain universal standards from regional business units that adapt products to local economic contexts and regulations. A monthly three-hour Technology Executive Board meeting with 20 senior leaders coordinates roadmaps and enforces standards across 23 countries to prevent redundant development.
• •Platform Economics of Security: Security costs function as fixed expenses that distribute across growing user bases, creating scale advantages. Moving from 200 million to 300 million consumers does not require 50% more security investment. This economic model enables purchasing leading-edge security tools and hiring top talent while maintaining profitability. Security spending represents the enabling cost for all other business investments, not a return-on-investment calculation.
• •Data Depersonalization Practice: Experian's credit scoring models analyze behavioral patterns without requiring personal identifiers like age, gender, ethnicity, or sexual orientation. The company traces this approach to founder C. Ramo, who tracked repayment patterns for pharmaceutical loans in 1800s England based solely on behavior, not demographics. This heritage informs current practices where most analytics use depersonalized data, with no plans to expose datasets to public AI models.