Credit reporting giant TransUnion LLC today is introducing a new artificial intelligence capability designed to accelerate complex credit analytics workflows and make them accessible to a broader range of users across the financial services sector.

The Chicago-based company’s AI Analytics Orchestrator Agent is built on top of its TruIQ software and the underlying OneTru data and analytics platform. It leverages Google LLC’s Vertex AI development platform and Gemini generative AI models.

The technology is designed to automate tasks that historically required large teams of data scientists and weeks or months of manual work.

“Those workflows take a lot of data science time, whether it is building a new model to underwrite something like an auto portfolio or thinking through a pricing model or analyzing why sales are being lost,” said Venkat Achanta, chief technology, data and analytics officer at TransUnion.

Financial institutions frequently rely on complex analytics projects to answer strategic questions about pricing, credit risk and product strategy. Achanta said these engagements often involve extensive data preparation and modeling work.

“There are a lot of what-if scenarios,” he said. “There are details of looking at their lending policy, their model, their outcomes and connecting it all.”

Governed workflows

TransUnion said the new agent automates much of that process by translating natural-language requests into governed analytical workflows that can generate models, studies or strategic recommendations. Many of the models are based on tried-and-true processes TransUnion developed internally.

“We’ve codified that effort into something that used to take weeks and months down to hours and minutes,” Achanta said.

The system analyzes requests, identifies relevant workflows and produces results while documenting the reasoning behind them. Audit trails are a key feature that TransUnion built in to satisfy regulatory demans.

“You can ask questions with natural language and in a very transparent way it understands the request, breaks it down and talks to you in terms of its internal logic,” Achanta said. The system can then “invoke governed auditable workflows and be get you thorough output, whether it’s a model deployment or a study.”

He emphasized that governance and transparency are essential because the analytics involved often affect core lending decisions.

“This is some of the most mission-critical work that a bank could do,” he said. “It gets to the core of their decision-making policy and it’s important to use an AI agent whose outputs are auditable and governable.”

The automated system generates “a big runbook of governance documentation,” Achanta said. “What are the key parameters? Why did it answer a question this way? What are the factors that led to that probabilistic answer?”

TransUnion said the capability reflects decades of institutional knowledge and experience analyzing credit data.

“We have over 800 data scientists who have done this for decades,” Achanta said. The agent “codifies our domain knowledge.”

Initially, the system is being used internally by TransUnion’s own analysts before being rolled out more broadly.

The company plans to make the capability available to external customers through TruIQ subscriptions after further testing and refinement. It also expects to expand the scope to marketing and fraud use cases. Though Google was chosen for its full-stack AI features, Achanta said the platform can work with other models as well.

“Our code platform is transportable to many foundation models on Google, but we can use Claude, Llama and others,” he said.

Photo: TransUnion