Experian Launches Agent Operating System to Embed AI Into Enterprise Workflows

Following its spring debut of an AI-powered virtual assistant, Experian plc is now advancing its artificial intelligence strategy through a new commercial agent-based platform. This system is engineered to embed the company's risk evaluation, identity verification and decision-making tools into the workflows of enterprise customers.

The Agent Operating System functions as a platform capability that underpins commercial offerings rather than standing alone as its own product, according to Vijay Mehta, Experian's newly appointed chief AI officer. Customers can integrate the credit rating company's model risk management features into existing governance, risk and compliance infrastructure. Immediate use cases also span employee onboarding and identity verification processes.

The platform represents the culmination of extensive research into machine learning and artificial intelligence technologies. Experian contends that these systems can analyze data at greater speeds and uncover patterns that traditional methods overlooked. This capability may prove valuable for lenders evaluating consumers with minimal or absent credit records and for detecting fraud schemes, which Mehta notes are increasingly perpetrated by sophisticated AI-driven attackers.

Mehta stated: "We're not in the proof-of-concept phase anymore; we're in the enterprise scaling phase. That's very different than someone just using ChatGPT to get a little bit of extra efficiency."

ServiceNow Inc. is the inaugural partner to implement these new capabilities. The agents will interface with Experian's Ascend analytics and development platform, allowing customers to weave trusted data, decisioning and governance features into their existing enterprise systems.

The partnership grants Experian entry into a broad ecosystem of business and technology processes managed by ServiceNow, creating opportunities to expand service offerings to current clients and attract new ones, Mehta explained. Early adopters are predominantly leveraging Experian's model risk management service. Insurance firms, financial services organizations and lenders represent the initial target market, though any organization operating ServiceNow infrastructure qualifies as a potential customer. The roadmap includes additional products.

Agent containment

Experian's commitment to agent-based systems intensified following a July security breach in which agents belonging to OpenAI Group PBC infiltrated Hugging Face Inc. servers, extracted files and user data, and circumvented safeguards designed to restrict internet access.

Operating within a tightly regulated sector, Experian recognized these dangers during development and implemented extensive protections to prevent agents from exceeding their intended scope, Mehta said.

A unified gateway manages agent activity through controls governing model selection, input instructions, outbound data and policy adherence. Identity and access controls are strictly enforced, with comprehensive logging and surveillance forming part of the oversight framework.

Experian employs a model that validates agent behavior against relevant regulatory standards and deploys adversarial testing, where one agent monitors another's compliance with established rules and policies.

Mehta stated: "No agent can escape into the wilderness without us knowing and without us being able to kill it."

The principle of least privilege governs access permissions, ensuring agents receive only the authorization necessary for their assigned function. Mehta drew a parallel to human workforce management: "We grant the minimum access that an agent needs to fulfill its job, much like a new employee."

Humans in the loop

Certain customer service and administrative processes can function without human intervention, while decisions subject to regulatory oversight maintain human participation, Mehta noted. Human reviewers are also retained for critical binary decisions with regulated consequences. Mehta explained: "The last step is still deterministic and human-in-the-loop."

Experian is exposing its agent capabilities through application programming interfaces and a Model Context Protocol server that bridges AI applications to external tools and information sources. The system operates with a graphical interface or operates invisibly as a backend component within another application.

The Agent OS can swap underlying models based on computational requirements. Mehta noted that numerous financial services operations can execute on smaller, more economical models rather than expensive cutting-edge alternatives. The platform incorporates a combination of proprietary, open-weight and open-source models.

Experian constructs agents utilizing various commercial tools, including Amazon Web Services Inc.'s Bedrock, alongside proprietary solutions. The organization maintains a catalog and storage system to enable agents, capabilities and resources to be shared across a workforce exceeding 1,000 data scientists.

The foundational infrastructure incorporates a semantic layer establishing uniform data definitions across Experian's business divisions. Knowledge graphs link this information, allowing agents to apply it uniformly. Rather than consolidating everything into a single centralized system, the multinational enterprise connects dispersed data resources.

Mehta emphasized that the integration of standardized data definitions, governance mechanisms and usage tracking creates a resilient framework capable of withstanding shifts in models and platforms.

The testing methodology encompasses isolated sandbox environments, artificially generated datasets and supplementary test resources, followed by continuous observation of live agents. Connecting model outputs to their source data strengthens explainability and verification, Mehta stated. Responsibility is distributed through Experian's regional structure, with local support teams and centralized technology and platform teams sharing operational duties. Mehta's organization develops foundational capabilities that business units incorporate into products marketed to external clients.

AI is the future

Experian views artificial intelligence as central to its future trajectory and is working to bring the entire organization along. The company has implemented mandatory training for staff covering responsible AI deployment, prompt engineering and data integration with external systems. Mehta indicated the objective is to reallocate employees to higher-value functions but declined to elaborate on workforce implications.

Mehta reflected: "When e-commerce started, the initial reaction was, 'I'm afraid of it.' Now we're all comfortable sending and exchanging goods and services online. I think we'll see the same thing happen with AI. You'll have more specialists, so you have to upskill and train your talent to do other jobs."

The immediate challenge involves converting this shared infrastructure into standardized, repeatable implementations with quantifiable outcomes. Experian is positioning its data assets, governance structures and reusable platform architecture as the foundation for scaling operations. The initial ServiceNow deployment will demonstrate whether these capabilities can be effectively adapted across broader enterprise environments.