From credit ratings to AI in the cloud: Experian’s quiet reinvention

It started as a note taped to the wall during a 2018 brainstorming session in Experian plc’s consumer business unit. The two-word idea, “Score Boost,” was one of 100 ideas that came out that day, but it was the one that helped save the company’s $2 billion consumer services business.

Experian’s consumer division was struggling at the time, under pressure from fintech startups such as Credit Karma LLC and posting multiple unprofitable quarters. In early 2019, it launched Experian Boost, a free service that allows consumers to add payments to their credit records that don’t normally appear on credit reports. Consumers who pay their bills on time can improve their credit scores by an average of 13 points. More than 17 million people currently use the Boost mobile app.

Today, Experian’s consumer business is not only profitable, but also central to the 125-year-old company’s larger transformation from a credit rating agency to an analytics and software provider. The company currently derives 35% of its global revenue from its software and platform. This change is being driven by Amazon Web Services Inc.’s massive move to the cloud and new efforts in artificial intelligence.

Cloud-first, AI everywhere

While Experian’s core business of providing reports on the creditworthiness of individuals and businesses remains important, the company is increasingly positioning itself as a provider of infrastructure for decision-making in financial services, from fraud detection to real-time risk assessment. We combine cloud services and AI to automate complex data migrations, minimize downtime, improve data accuracy, and provide more scalable and secure services.

This is in line with broader industry changes. Like most financial data companies, Experian is racing to stay ahead in a cloud-native world where data flows continuously and customer expectations are higher than ever.

Alex Lintner led the team that developed Boost. Since then, he has held a wide-ranging role as Experian’s Chief Executive Officer of Technology, Software Solutions and Innovation, overseeing North American operations and developing global strategy. In an interview with SiliconANGLE, he explained how Boost is part of a larger story about how Experian is reimagining its role from a scorekeeper to one that enables smarter financial decisions.

The company’s transformation was fueled by a 10-year agreement with AWS that “enhanced performance, scalability, and reliability, reduced operational costs, and strengthened security,” Lintner said.

The cloud also allows Experian to consolidate its internal data into a data lake that is expected to grow to more than 100 petabytes within 18 months. We built our own tools to manage migrations, including software to load and reload data into AWS S3 buckets with real-time updates. “Our customers tell us we have the most up-to-date data,” Lintner said. “Not all stations can say that.”

At the heart of Experian’s AI operations is Ascend, a platform that began as a sandbox for data scientists to visualize data and test models. This led to “AscendOps”. It automates the handoff between data science and information technology departments, turning models built in Python or R into production-ready code for financial institutions.

“This process used to take three weeks,” Lintner said. “Now it takes two to three days, but it still requires human quality control.”

Reducing fraud

Experian has developed an AI-powered fraud detection model that it says dramatically outperforms traditional rules-based systems. Detection rates improved by 37% for loan scenarios and 45% for credit card applications. Lintner said one financial institution avoided losses of more than $250,000 in one fraud attack cycle. Meanwhile, the amount of manual reviews was cut in half, improving the customer experience with fewer false positives.

The company is also investing heavily in AI agents that can operate semi-autonomously and perform actions. The digital agent monitors the health and accuracy of the model and directs data scientists when anomalies are detected, suggesting new data sources to improve the model’s performance.

Like many companies experimenting with agents, Experian tests them extensively and selects the most effective ones. “We’ve built hundreds of agents, but most of the usage is in the top 20 companies,” Lintner said.

One popular agent focuses on “model drift,” alerting teams when a model’s performance begins to degrade and recommending changes that can be made via a drag-and-drop interface. “Previously, we had to manually rebuild the model,” Lintner said. “Now you just accept or decline the prompt.”

The AI ​​system monitors for suspicious behavior patterns, such as a server initiating tens of thousands of loan applications. The company uses behavioral biometrics, such as typing speed and frequency of typos, to determine whether a person is real or a bot. Near real-time updates detect fraud faster than ever before. “When bad actors develop new strategies, you need to react quickly,” Lintner said.

AI is also helping fight the “Frankenstein ID.” Frankenstein ID is a tactic in which fraudsters construct a fake ID using real and unrelated data, usually obtained from different sources. Lintner said his company’s systems can now detect these fake documents much more quickly than before.

“The goal is to minimize the reach, whether that means catching fraud before it happens or limiting the number of accounts affected,” Lintner said.

LLM for Compliance

Beyond fraud detection, Experian has built extensive language models tailored for regulatory compliance. One early application is automating the production of reports required under the Supervisory Guidance for Model Risk Management, a set of principles known as SR 11-7 required by the Federal Reserve System for validating credit risk models.

Financial institutions once spent months creating SR 11-7 reports that exceeded 200 pages, often requiring entire data science teams to spend the last few months of the year. Today, LLM automatically generates the necessary documentation from model metadata and other sources, checks its completeness, and outputs a PDF for regulatory authorities. “Before, we had 30 people just doing documentation,” Lintner says. “You can now auto-generate using LLM.”

Technology is not the only driver of change. Lintner acknowledged that educating Experian’s 23,000 employees on AI tools is an ongoing process. “There are 11,000 people using it at a proficient level,” he said. “It depends on what kind of work they do.”

The company tracks adoption using what it calls “love metrics,” which measure tool usage time, repeat usage, and organic growth. “We don’t promote these tools,” Lintner said. “If usage increases, we think it’s worth it.”

From helping consumers improve their credit scores to providing banks with real-time insights, the company’s AI shift is ultimately about expanding access and trust. As Lintner says, the goal is not just to make the system faster, but to make smarter, fairer financial decisions.

Photo: Experian