How Experian pivoted to responsible AI and grew its business in the process

Credit bureau giant Experian has undertaken an ambitious cloud migration to support new AI services that generate 35% of its global revenue. Moving from on-prem infrastructure to the cloud made it easier to automate complex processes, reduce downtime, improve data accuracy and reduce costs.

This took years of careful planning, according to Alex Lintner, CEO of Software and Technology at Experian. The firm's infrastructure previously operated across several regional data centers. It has already helped lower management costs by nearly 40%. Now Experian is using some of those savings to invest in technology to assist customers on their AI journeys as well.

Centralizing systems made it possible to develop, train, test, and deploy both generative and agentic AI models globally within days rather than months. It also made it easier to democratize safe innovation across all of the core businesses. For example, business units are using the platform to develop and improve domain-specific small language models. This infrastructure has also helped the organization to roll out new products and services for real-time modeling and decision-making for financial institutions, with built-in compliance enforcement. 

Lintner explains:

Our core business of reporting on creditworthiness remains a vitally important service to our customers and consumers. But for almost a decade now, Experian’s role has expanded far beyond that. Today, we provide the infrastructure that enables smarter, faster, and more equitable credit and fraud analytics, risk evaluation, and decision-making across nearly every financial services function. Our solutions help institutions manage credit, protect identity, detect fraud, uncover business opportunities, and assess regulatory risk in real time.

Over the past decade, these new AI products and services have evolved to assist, drive, or eliminate some of the tedious, time-consuming and error-prone workflows for its financial industry customers. And this has been good for the bottom line - software and platforms now represent 35% of revenues, up from roughly ten percent just five years ago. 

A ten-year strategic partnership with AWS has been critical to this evolution. Lintner says this has provided the network performance, scalability, global reach, and highest standard of security to operate globally, while lowering operating costs: 

Most of Experian is now a cloud-first organization that moves faster, innovates responsibly, and helps the financial sector navigate the need to build consumer trust through transparency.

Platforms on platforms

One example of this has been the rollout of a new Experian platform that runs on AWS. This has enabled their team to focus on their essential differentiation rather than just re-inventing the infrastructure stack. The early focus was to provide a sandbox that allowed data scientists to explore and test financial models. This has since grown into the center of Experian’s AI operations, where financial institutions use it to analyze data in real time, create and deploy models, and turn insights into action more quickly.

A Forrester study found that these financial institutions could achieve a 180% ROI within three years by improving their ability to predict the risks of new financial products. As a result, they could bring these products and services to market faster. Lintner explains that each new version of their platform provides the foundation for the next one: 

The second version seamlessly integrated models into production systems, the third added advanced fraud modeling, and the fourth introduced agentic AI. These agentic digital assistants monitor model drift, suggest recalibrations, and automatically prepare regulatory documentation. The result is up to 90% less manual work and faster, more accurate compliance reporting.

This iterative approach has also helped Experian improve its fraud prevention models to keep pace with the bad guys. These new AI-powered variants can adapt to new fraud schemes more effectively than simply adding new rules to traditional rule-based approaches. This is especially important as bad actors increasingly employ AI bots to rapidly develop and roll out new attacks. 

For example, Experian is now analyzing behavioral biometrics, such as typing speed, mouse movement, and error patterns, to distinguish genuine users from automated bots. AI systems can also improve monitoring for unusual activity, such as sudden surges in application volume, and alert fraud analysts more quickly. In one case, a lender prevented more than $250,000 in fraud during a single attack cycle.

Transition challenges

Lintner says most of the transformation challenges involved the management of the transition period:

Unfortunately, there is not a magic switch that can be flipped to instantaneously place all our applications at all our customers on the cloud. We must work application by application, client by client. It’s a marathon, not a sprint.

During this hybrid state, it was important to invest in tooling and rehearse failover practices. Lintner was concerned that the transition would become a source of customer challenges.  

It was also important to develop a new way of thinking about collaboration, development, data ownership, and risk to strike a careful balance between speed, reliability, and compliance. Lintner advises other organizations beginning a similar journey to treat transformation as a strategic business initiative rather than an IT project. This included defining measurable goals, building a strong governance framework from the start, and investing in employee readiness. For example, Experian has launched a program to train employees on the responsible use of AI and how to avoid increasingly sophisticated fraud attacks. 

As a result, generative and agentic AI are widely used across nearly every function at Experian, including data analytics, modeling, decisioning, software development, and fraud operations. It also uses these tools to accelerate coding, personalize customer experiences, and model complex fraud scenarios. Lintner says this internal experience has also helped shape better conversations with customers:

An important philosophical point for Experian is that we use AI internally first, learn from those experiences, then extrapolate best practices and process improvements for our customer solutions. In other words, we experiment on ourselves first to work out the kinks in whatever we propose to our customers.

Creating a foundation for responsible AI

There has been a growing focus on the different ways in which agentic AI introduces new concerns about responsible AI. It is no longer enough to understand how models arrive at predictions. Enterprises must also find better ways to monitor the actions they take in production. To tackle this, Experian has established a Responsible AI council to collaborate on ethics, risk, inclusion, and compliance. One aspect of this involved extending the use of observability tooling to trace model lineage, assess bias, and measure equity. This ensures that AI-powered decisions remain transparent, explainable and accountable. 

Lintner predicts that this focus on responsible AI will help Experian prepare for a coming wave of intelligent financial assistants. He believes these AI-driven systems will help consumers and institutions make informed decisions by providing real-time insights, detecting fraud, and recommending personalized actions. The goal is to deliver reasoning systems that can learn from context and take responsible action within defined limits. He says:

Ultimately, our aim is to create a connected financial eco-system where trust and transparency guide every interaction.

This goal is driving investments in privacy-preserving AI, secure data-sharing frameworks, and governance standards that protect consumers while promoting innovation. Lintner argues this will give individuals greater control over their information, help leaders make fairer decisions, and work with regulators to build confidence as a positive force for inclusion and trust. 

My take

Lintner paints a bold vision for the future of responsible AI that will appeal to regulators and financial customers. For sure, it promises a viable middle ground alternative to the move-fast-and-break-things approach that seems to be emanating from Silicon Valley innovators.  

However, building real trust with consumers will require a greater emphasis on treating consumers as partners rather than subjects. But then, who is going to pay for it?  If it’s the financial institutions, it seems likely that consumer interest will take a back seat to revenues and profits. At least in the US, there does not seem to be a pragmatic path that strikes a more equitable balance between financial institutions, data stewards and consumers, particularly in the near future.  

I, for one, succumb to the fantasy that there are even bigger opportunities for businesses and consumers in the future as outlined in Tim Berners-Lee’s Unfinished Story of the World Wide Web. But then there is far too much money being made today on legally re-packaging and re-selling our data today.