How LLMs and AI Agents Transform Unstructured Data Analysis

Posted June 10, 2025

In a recent episode of AI + a16z, Instabase founder and CEO Anant Bhardwaj joined Guido Appenzeller, a16z Infra partner, to discuss the transformative impact of Large Language Models (LLMs) on unstructured data analysis. The conversation highlighted real-world applications, such as enabling banks to verify identities and approve loans via WhatsApp, and explored the future potential of AI agents in automating workflows.

Key Discussion Points

• Legacy RPA Limitations: Traditional robotic process automation (RPA) struggles with unstructured inputs like PDFs, emails, and handwritten documents.
• Layout-Aware Models: Instabase developed advanced models to extract insights from complex documents by understanding their layout and context.
• Predictability Over Perfection: For enterprise AI solutions, consistency and reliability are more critical than achieving flawless results.
• AI Agents at Compile Time: The growing role of AI agents in designing and configuring systems before they run, rather than just during execution.
• Decentralized AI Systems: A vision for federated AI systems that can scale automation across intricate workflows while maintaining data privacy.

Contributors

• 
  Anant Bhardwaj
  Founder and CEO of Instabase.

• 
  Guido Appenzeller
  Investor at Andreessen Horowitz, focusing on AI, infrastructure, and open-source technology.
  Follow on X | LinkedIn

Related Content

• To Help AI Startups Compete, Use Revenue-based Thresholds
• Who’s Coding Now? AI and the Future of Software Development
• What Is an AI Agent?
• DeepSeek, Reasoning Models, and the Future of LLMs
• Data Management for Enterprise LLMs

This discussion underscores the rapid evolution of AI technologies in unlocking the value of unstructured data, paving the way for more intelligent and autonomous systems in the enterprise.