Experian built an AI tool that allows data scientists to interact with Experian’s data assets using natural language. In a way that wasn’t possible before, it puts deep data insights in users’ hands more intuitively, without a heavy reliance on documentation.
In the second part of a special episode of The Disruption Is Now, recorded at Money2020, host Greg Matusky sits down with Experian’s EVP Shri Santhanam to discuss the company’s leap forward and what it means for workflows, decision-making, and compliance. The conversation reveals how Experian has created systems and structures throughout its organization to make the most of AI.
Watch now:
Key takeaways
Experian’s AI assistant acts as a 24/7 data expert
Experian introduced an AI assistant that functions like a personal data scientist available around the clock. This assistant understands every data attribute and is knowledgeable about regulatory constraints.
Clients can interact with their data using natural language, asking questions directly within their development environment. For example, a data scientist working on a specific attribute can ask the assistant about its permissible uses, compliance considerations, and even get code snippets to implement it effectively.
Shri described it as having one of Experian’s best data scientists at your fingertips, reducing the need for extensive documentation or expert consultation.
AI can turn unstructured data into accessible knowledge
The assistant bridges the gap between structured data (like credit attributes) and unstructured data (like documentation and regulatory guidelines).
Data scientists often struggle to understand what thousands of data columns represent and how they can be used legally. The assistant taps into unstructured knowledge, providing context and explanations for each data attribute. This saves time and helps clients use data more effectively and responsibly.
An AI risk council ensures compliance
Given the regulated nature of the financial industry, Experian prioritized compliance and explainability in developing the assistant.
They established an AI Risk Council consisting of technologists, legal experts, compliance officers, and AI specialists. This council ensured the assistant met all regulatory requirements and minimized risks like AI “hallucinations” by carefully curating data sources. They tested the assistant across 28 to 30 specific skills, ensuring accuracy and reliability before deployment.
Collaboration speeds up AI development
Experian fostered collaboration across departments, breaking down silos between technologists, legal teams, and compliance officers. This multidisciplinary approach accelerated the assistant’s development and ensured it met diverse requirements.
Shri shared that by having experts educate each other, they accelerated learning and created a “superhighway for POCs to scale,” moving projects from proofs of concept to production efficiently.
In addition, Experian encouraged its 20,000 employees to engage with generative AI. Recognizing that great ideas can come from anywhere, this grassroots approach led to unexpected insights and use cases, fueling the organization’s AI journey.
Key moments
- Experian’s generative AI assistant overview (1:16)
- Reinventing how financial services operates (3:08)
- How to access less accessible data (6:10)
- Working with natural language to understand data (8:42)
- Focus on AI explainability and reducing hallucinations (10:24)
- The role of AI councils and collaboration (13:18)
- AI as the nexus of disciplines (19:46)
- Two key principles regarding AI at Experian (23:22)
- Shri’s son using ChatGPT (25:28)
- Creating the right apparatus to move from prototypes to production (27:57)
- What has surprised Shri the most about the trajectory of generative AI (30:19)
- The most outrageous future we can’t imagine (37:22)
