Every Agent Needs a Box — Aaron Levie, Box

Agent Data Management: AI agents require a secure and governed platform (like Box) to access and manage enterprise files, including permissions and collaboration, fundamentally transforming how organizations utilize their data.

Enterprise Adoption Hurdles: Unlike AI coding, general knowledge work faces major obstacles such as complex access controls, diverse data formats, poor documentation, and the high-risk "slop" of agent-generated content, necessitating a multi-year workflow re-engineering effort.

Context Engineering & Data Quality: Current AI models struggle with vast, messy data. Effective agent performance depends on improved search systems, better model judgment (knowing when to stop searching), and companies adopting superior documentation practices to provide clean, relevant context.

Agent Identity & Liability: A critical unsolved problem is defining agent identities and assigning liability for their actions, as agents cannot simply be treated like human users due to oversight and privacy concerns.

Future of Work: The rise of agents will lead to an order of magnitude more agents than people, creating a massive infrastructure opportunity for data management and making roles like Developer Relations (DevRel) increasingly vital for guiding agents to use specific services and APIs.