Building an Agentic Platform — Ben Kus, CTO Box

Ben Kus, CTO of Box, shares Box's AI journey, focusing on data extraction from unstructured content for enterprise clients. He highlights the shift from traditional ML models to generative AI, which initially showed great promise but struggled with complex documents and accuracy. Kus details how adopting an agentic approach, using AI agents with directed graphs, significantly improved accuracy and flexibility. This architecture allowed for continuous improvement by tweaking prompts and adding double checks, rather than complete redesigns. The agentic foundation also enabled new capabilities like deep research on user data within Box. Kus emphasizes the importance of separating agentic framework development from scaling generic processes and encourages an AI-first mindset for teams.