TLDR
- TransUnion’s new agentic AI system can multiply its data science teams’ output by up to 5× without hiring more people.
- Analyses that once took weeks can now be done in hours, enabling faster lending and pricing decisions.
- TransUnion’s consulting services can now be amplified through agentic AI automation.
TransUnion has introduced an agentic AI system designed to increase the output of its data science teams by three- to five-fold while extending its analytics services to more clients − without adding headcount.
The credit reporting and analytics giant has built its ‘AI Analytics Orchestrator Agent,’ which the company said compresses complex analytical work that once took weeks into hours or minutes, while keeping humans in the loop. The system is being used internally and should be rolled out to clients in coming months.
Data analysis is one of the company’s core services: helping financial services firms make decisions about lending, pricing and market strategy using data and custom-built models. The agentic AI system will enable TransUnion’s data scientists to perform more work faster for its customers.
“The demand for our data scientists is huge,” said Venkat Achanta, the chief technology, data and analytics officer at TransUnion, in an interview with The AI Innovator. “The more we engage with our customers, the more depth of our understanding of our customers increases with the product and data. So, we want to offer more of this; this is a way to scale out that function.”
TransUnion employs more than 800 data scientists who work directly with clients to build models and analyze business problems. The demand for that work has grown faster than the company can staff.
“We could never add enough people to provide more of this service, because the more we expand the service, it’s linearly to the number of people,” Achanta said.
The orchestrator agent is designed to break that constraint. Instead of requiring a team of data scientists to run each analysis, the system allows those workflows to be executed through a conversational interface. Under the hood, it connects natural language inputs to structured analytical processes built from TransUnion’s historical models and domain expertise.
“Think of this agent as using AI and the natural language model of LLMs, but in a very governed, auditable way,” Achanta said.
The company developed the system on its OneTru platform and integrated it with tools from Google Cloud, including Gemini models.
The design reflects a broader shift in enterprise AI. Rather than replacing human decision-making, companies are using AI to scale specialized knowledge that was previously difficult to replicate.
From bespoke consulting to repeatable workflows
Before the new system, TransUnion’s work with clients often resembled consulting engagements.
Banks would bring strategic questions – such as how to enter a new market or why loan conversions were declining – and TransUnion’s teams would analyze the problem using proprietary data and custom models.
“It takes our data scientists engaging with them to be able to take their strategic problem … and build custom insights,” Achanta said.
Those engagements typically took weeks and involved extensive documentation to satisfy compliance requirements. Each analysis had to be explained, tested and validated.
The orchestrator agent does not eliminate that process. It encodes it. “What this is, is essentially these workflows were done earlier,” Achanta said. “But now the time compression is the biggest thing.”
Tasks such as lost-sales analysis or market entry modeling can now be completed in hours instead of weeks, he said.
The change is as much about consistency as speed. By standardizing workflows, the company can deliver the same type of analysis repeatedly, rather than rebuilding it for each client.
That shift turns what was once a bespoke service into something closer to a platform.
How the system changes decision-making
The practical effect for lenders is faster access to insights that were previously constrained by time and expertise.
A bank, for example, could ask why it is losing loan applications at a certain price point. The system would analyze approval models, competitor behavior and potential pricing adjustments, then simulate outcomes across different scenarios.
“This entire analysis can be done – and can be shown to them on a spectrum of where they can be on the price point, and what they can accomplish at each price point,” he added.
The system draws on large volumes of proprietary data and prebuilt workflows. It uses a combination of machine learning and rule-based reasoning to ensure results are grounded in established methods.
Achanta described the approach as “neuro-symbolic,” combining neural networks that let the LLM understand the user’s query in natural language with structured logic based on predefined rules, models and workflows.
That structure is critical in a highly regulated industry. “The most important thing is it’s a very transparent, auditable, governed workflow but using a very agentic approach with a chatbot-style front end,” he said.
According to the company, the system breaks down each step of an analysis and explains it in plain language, allowing users to understand how conclusions were reached.
This addresses a common concern with generative AI: producing answers without showing its reasoning. Explainability is critical in credit decisioning. Lenders must be able to explain how models operate, test for bias and document their processes for regulators. The orchestrator agent is designed to generate that documentation automatically and allow users to query it.
“Your compliance person in the bank can ask, ‘how was this model built? What is the bias? How is the confidence? How did you test this?’” Achanta said. “There is a full set of governance documentation, but if you don’t care to read about it, you can have a question-and-answer session.”
A shift in how data companies compete
The system serves financial services firms of all sizes, including many mid-sized banks, regional lenders and fintech companies leveraging the service. While large global banks often build their own models and infrastructure, they still use TransUnion’s data and may collaborate on specific problems.
The focus is on lending, where decisions about risk, pricing and customer targeting are central to profitability. “Think of it as a new model for lending,” Achanta said.
The initial rollout is internal. TransUnion’s data scientists are using the system first, allowing the company to test its accuracy and refine workflows. The next step is to make it available through the TruIQ platform, where customers will be able to access it more directly.
“Today, our data scientists are using this agent to be able to do more of this custom service. We are going to bring this to our subscribed customers on our platform very soon,” Achanta said. “This is the tip of the iceberg.”
By embedding analytics and decision-making into a platform and supercharging those capabilities through agentic AI, the company is adding more heft to becoming a provider of outcomes rather than just inputs.
“This is a way to scale out that function,” he said.
TransUnion also is framing the technology less as a replacement for its data scientists but to make each of them count for more.
“The flywheel has already been proven when our data scientists engage with the customers. Not only are we solving a specific problem for them, but we are helping them with the depth of understanding of all we provide, and creates a nice flywheel of consumption as well,” Achanta said.
