TransUnion’s AI initiatives are paying off in unexpected ways

From launch a massive data lake Earlier this year, to support its analytics, application development, and artificial intelligence use cases worldwide, TransUnion LLC began realizing the benefits of deploying artificial intelligence models in back-office cases – and they’re not necessarily what you might expect.

Self-service AI models for data preparation help break down silos that have formed around areas of expertise. Instead of throwing problems over the wall and letting them be solved by specialists, human analysts are now turning to AI to find solutions. This eliminates blockages in workflows and makes work more meaningful, according to Venkat Achanta (pictured), TransUnion’s head of technology, data and analytics.

“Users can now orchestrate end-to-end processes and eliminate silos,” he said. “What they previously had to ask another group to do, they can now do themselves. An AI-empowered person can accomplish a much more complex set of tasks.

The result is that skills stocks move up the hierarchy. “People who weren’t able to analyze data are now becoming super-analysts,” he said. “Data analysts move up the data ladder to become modelers and take on higher-order tasks. This pushes everyone to move up the ladder. »

As a highly regulated company, TransUnion AI focused most of its early AI deployments on internal functions such as reducing fraud, data loading, and improving developer productivity. The results have been generally positive and, in some cases, spectacular.

Force multiplier

Large language models have improved fraud detection by simulating investigative tasks that previously required human agents.

“Agents need to understand the grammar of fraud,” Achanta said. “A certain sequence of steps can mean one thing, and a pause between steps can mean something else. If you think about how LLMs work, you can create a deep learning model tailored to specific cases of financial fraud, and we find that some of these techniques apply very well.

AI has improved fraud detection accuracy by 40% and reduced false fraud alerts by more than half. “These two factors work as a force multiplier,” Achanta said.

AI is also proving to be a valuable partner for data integration by making decisions that previously required human review. Routine tasks such as data validation can be distributed more quickly and accurately by machine. “It recognizes data elements, labels them and applies the compliance rules that you have so users don’t have to sit there and map each one,” he said.

For example, a model might be asked to generate an alert for a phone number with no area code but pass along a four-digit Social Security number because “that’s all we need in a privacy-friendly world.” “, he said. “What he can do upfront is incredible. Performing these data quality checks after the fact is very expensive.

In software development, LLM-based code wizards are part of a broad modernization program to convert aging .NET code to Java. “There are tremendous benefits in code conversion when you apply AI,” Achanta said.

Jobs are evolving under the influence of automation, which naturally arouses some apprehension. Allaying people’s fears “is a work in progress,” Achanta said. One of the most effective ways he has found to combat anxiety is to focus on the quality of work.

“For a developer, this means generating the task case rather than writing the task case code,” he said. “These are trivial things that developers don’t want to do. People achieve more meaning by completing more creative tasks.

Photo: SiliconANGLE