Experian’s Chief AI Officer on the Next Stage of AI Deployment

TLDR

• Experian was early to the generative AI trend and now it is shifting its focus from AI adoption to redesigning workflows around agents.
• CAIO Vijay Mehta said the company is seeing real ROI from AI.
• Experian built guardrails that can stop an AI agent dead in its tracks, if necessary.

Vijay Mehta, Experian’s new chief AI officer, has spent much of his career immersed in technology, data and analytics – a world of models, numbers and code that can be clinical and cold.

But ask him what he loves about AI, and he gets passionate.

“I’m really excited about the future because when I think of the benefits to consumers, to businesses, to humanity as a whole, we are just starting to scratch the surface,” Mehta told The AI Innovator in an interview.

“Even beyond financial services, the things that can happen with AI – finding new genes and finding new proteins, curing cancer – that’s how big this is and that’s amazing,” he said. “Wow. Really, that’s the art of the possible here.”

Mehta brings that passion to his work. At Experian, he has been spearheading AI initiatives for years. This latest promotion makes it official.

“My role is to ensure that we’ve got a common strategy and direction around AI for the organization,” Mehta said. “We are a big company, and we are broken up by many business units. My job is to make sure that we are all marching in the same direction, that we are accelerating our agenda both internally and externally.”

The other parts of his job are to manage risk, security and compliance as well as ensuring Experian has the right partners and capabilities. “All of that comes together into the AI office,” he said.

Experian has already moved well beyond the proof-of-concept stage, Mehta said, with AI capabilities running in production throughout the company. He said Experian is now transitioning from broad AI adoption toward what he calls “reimagination” – redesigning how work itself gets done.

Instead of simply giving an employee a chatbot or coding assistant to perform an existing task faster, for example, the company is examining processes to determine where work can be delegated to an agent – or several agents working together.

“It’s that shift of mindset from where AI was an accelerator, or something that somebody can kind of bolt on,” toward “recreating the way that the actual work is done and being able to delegate to an agent and have the agent make decisions,” Mehta said. “That’s the journey we are on.”

What Experian has done with AI

The company, a global data giant and one of the Big Three credit bureaus, dove into generative AI early.

Within months of ChatGPT’s November 2022 debut, Experian began laying the groundwork for generative AI across the company. It trained thousands of employees and put a set of guardrails and policies around AI use. Traffic from these tools was routed to a control plane that gives the company observability, monitoring and security controls. Today, 85% of employees are actively using those tools, Mehta said.

It developed a GenAI risk framework and built the Experian GenAI Platform, an internal ecosystem designed to support both employee and customer-facing applications and allow secure use of company data, according to its annual report for the 12 months ended March 31, 2024.

Experian also started rolling out a code-development assistant to its large developer workforce and other productivity tools across the organization, while incorporating its own Experian GPT into products including a digital financial assistant and its Ascend platform, the report said.

The work has since broadened. Internally, Experian has expanded AI across software development, analytics and business operations. It created a GenAI Center of Expertise, which Mehta said is now under his AI office.

It infused employee training and career development with AI. Its ‘MyCareer’ platform recommends learning and career opportunities and helps managers identify skills gaps. Nadia, its internal AI coach, helps employees set goals, build development plans and prepare for performance discussions while giving leaders personalized coaching and 360-degree feedback.

By the first half of fiscal 2026, 88% of developers were using coding assistants and employees created more than 6,000 custom GPTs.

Externally, Experian has moved into agent-based systems as well, launching an Agent Operating System to let its own, client and partner agents work together across lending workflows and Agent Trust to verify that an AI agent is actually connected to the consumer it claims to represent.

Most workloads don’t need the largest models

Behind those applications is an architecture designed to give Experian more control over which models employees and applications use.

Traffic from external AI tools and Experian’s own systems passes through a control plane that provides monitoring, security and observability, according to Mehta. The company can then route workloads to different models based on the task, language, region and regulatory requirements.

Experian uses a mixture of open-weight models it has adapted for its own environment and frontier models accessed through infrastructure such as AWS Bedrock.

The approach runs counter to the tendency of companies to reach automatically for the most powerful – and often most expensive – model available.

“Not everybody needs the most advanced frontier model,” Mehta said. “Most workloads can be done with a free or very low-cost model.”

He compared indiscriminate frontier-model use to using a sledgehammer for a job that requires a much smaller tool.

Notably, Mehta said Experian is already seeing real ROI from its AI efforts. He said the company measures returns through time savings, time to value, quality, velocity and traditional before-and-after financial metrics rather than relying on adoption or token use alone.

About those recent hacks

Recent incidents involving AI models from OpenAI, Anthropic and Google hacking external websites or bypassing intended controls during safety and cybersecurity testing have heightened concerns among enterprises about how much autonomy agents should have.

Mehta said Experian’s approach is to design those controls into the architecture from the beginning rather than add them later.

“When I look at our environment – and I’ll just say our environment internally and then what we provide to our customer base – there is no way that an agent is going to break out because we’ve got the guardrails on it at all the points of the process,” he said.

Experian’s Agent Operating System designed around a control plane would sandbox, monitor and stop an agentic process when necessary, Mehta said.

“That control plane, setting the guardrails, making sure that we’ve got the right monitoring, and the ability to kill a process – the ability to kill a process, I will repeat that – we’ve got all of that built in,” he said.

He said Experian’s risk profile also differs from frontier AI labs because its systems operate over a defined corpus of financial and identity data rather than giving a model broad access to the internet or scientific datasets.

“We start with our corpus of data and we put controls around that,” Mehta said. “And we know exactly how the data is moving.”

For companies still early in their AI deployments, Mehta’s advice is to pay attention to the basics: train employees, establish guardrails before deployment and route AI traffic through a controlled gateway where usage can be observed and policies enforced.

“Set those guardrails up front,” he said. “So much of safety and responsibility comes from the person interacting with the system.”

AI becomes the new middleman

As Experian transforms itself, it has to manage a world where AI assistants are inserting themselves between the company and its customers.

On the consumer side, Experian has increasingly put its services inside third-party AI platforms. It recently became the first personal finance app integrated with Google’s Gemini, allowing U.S. consumers who choose to connect their Experian accounts to ask Gemini about their FICO credit rating and factors affecting it.

The company has taken a similar approach with OpenAI. Experian has launched personal loan shopping, credit card shopping and an insurance marketplace inside ChatGPT, among other experiences.

The strategy reflects a potentially significant shift. Consumers who once went directly to Experian or a bank can increasingly begin their financial journey inside an AI assistant.

Mehta said Experian expects multiple interfaces to coexist.

“Multiple experience layers and multiple experience environments are the wave of the future,” he said, describing the strategy as meeting consumers and businesses “where they want to play.”

At the same time, AI agents are appearing on the other side of Experian’s business: inside the enterprises that use its data.

ServiceNow became the first integration partner of Experian’s Agent Operating System. Through a multiyear partnership, ServiceNow agents can tap Experian’s Ascend data and decisioning capabilities from within enterprise workflows.

That means AI is increasingly becoming an intermediary on both sides of Experian’s traditional business: between Experian and consumers, and between Experian and its enterprise customers.

Mehta argues that does not weaken Experian’s position. He believes it increases the importance of the data underneath those agents.

Asked whether an AI agent could eventually bypass Experian and obtain the information it needs elsewhere, he said it can’t reproduce their data. The company has decades of longitudinal financial and identity information whose provenance and regulatory standing matter when the output is being used to make consequential decisions, Mehta said.

“In order to be able to make a decision, you’ve got to have that trusted data source,” he said.

For Mehta, that mix of ambition and guardrails is central to the advancement of the technology. It’s still very early days, and AI has already proven to be a game changer in many domains.

“This is so groundbreaking and it’s moving so fast,” he said. “The speed of change is just so exciting. I’m so happy to be at the center of it for the organization and helping us navigate the times ahead.”