A pedigree in information collection that spans some two centuries is not enough these days, given the expectations of speed and innovation in the AI era. That is one of the reasons why data, analytics and consumer credit reporting giant Experian worked with agentic AI startup Skyfire and other startups to develop a framework to verify that AI agents behave as instructed.
Kathleen Peters, Experian’s chief innovation officer, spoke with InformationWeek about how her company taps startups to identify new opportunities in the AI age. That includes the partnership with Skyfire to create the Experian Agent Trust framework to validate transactions conducted by AI agents.
The framework includes resources from Cloudflare and Visa Intelligent Commerce, with Skyfire providing a standardized method for agent-related information exchange across platforms.
For its part, Experian has extensive experience understanding consumer and business identities and verifying them deterministically for compliance and probabilistically for fraud detection or fraud risk, Peters said. The company’s history stretches back to 1826 and has seen its share of change. The ascension of AI, and now AI agents, is part of a generational technology evolution in the transaction space.
“The landscape is ever-changing from in-person to digital, and now to agents who are acting on humans’ behalf,” she said.
Peters has been with Experian for 13 years, spending the majority of that time in identity verification and fraud detection. Experian has been looking at the AI landscape, how CIOs are thinking about it from a workforce standpoint, and how it relates to identity access management, Peters said. This includes confirming that the people accessing these systems or seeking these services are who they say they are.
As more consumers use AI tools — including agents — in the purchase process, and companies such as OpenAI and Stripe enter into partnerships, questions arise about the leeway granted to those agents.
“Who am I giving this information to? Is this agent going to go rogue? The last thing I need to be doing is calling my bank and my credit card company, saying, ‘I don’t recognize these transactions, I don’t know how it happened,'” Peters said.
Verifying AI agents
Given that fraudulent transactions and billing errors are nothing new, there can be a lack of consumer and merchant trust in AI-driven transactions, she said. That is one of the reasons why retailers include security layers, such as firewalls, on their websites. Many of those measures are meant to protect against intrusion, such as bots, but the spread of AI has cast bots in a different light.
“Before AI, bots equaled bad. If there’s bots coming to your site, you wanted to block them because it’s not going to be something good. Either it’s an attack, it might be scrapers, who knows what it is,” Peters said.
Verifying the bot-like actions of AI agents is where Experian and its collaborators aim to come in. The framework Experian developed with Skyfire’s input is a way to “know your agent” as it acts on a customer’s behalf. “It’s like KYC [know your customer], which we are very familiar with, especially in AML [anti-money laundering] worlds,” Peters said. “We’re going to use the same approved methods that we do for traditional KYC today, especially in digital banking onboarding.” This can include requiring a credit card to verify a transaction.
Experian explores strategic investments in startups
Skyfire is an AI startup out of San Francisco that authored an open standard called KYAPay, Peters said. With KYAPay, developers can verify the identity of their AI agents and track agent activity. Working with Skyfire is part of Experian’s overall efforts to explore strategic venture investment opportunities, including M&As, she said. ” We’re excited about what’s on the bleeding edge in terms of capability.”
With the AI space growing, moving quickly and churning in recent years, Peters said Experian conducts in-house demo days to understand what startups are creating. “My team goes to the shows, does the desk research, meets with companies, goes to the incubators and curates a list [of startups to watch]. We do this about three or four times a year,” she said.
The demo days are conducted virtually, with pitches heard by attendees from different functions across Experian. Then Peters’ team divides and conquers — following up themselves or directing another Experian business unit to pursue collaboration with the startup. “That’s how we found Skyfire,” Peters said.
What stood out about Skyfire was its specialization in e-commerce and agentic commerce, along with the open standard for identity that they authored, she said. “We knew they’d be a valuable partner in helping us navigate the [agentic commerce] ecosystems. They’ve got an early read on where a lot of pain points and challenges are.”
Being a data and analytics company, Experian has used machine learning and AI techniques for a long time. The company’s research lab worked on neural networks for more than a decade, Peters said. As generative AI came to the fore, natural curiosity bubbled up inside the company, she said. “We put a risk council in place, we put guardrails in place, enterprise tools and we started looking for what would be the best AI stack within our company.”
Peters said Experian continues to look for technology developed by startups that could be applicable or adjacent to operations and services that the company wants to expand. “We may partner with them. We may make a strategic investment in them. We may put them on a watch list for acquisition,” she said.
