How Enterprises Can Move from AI-Assisted to AI-Native Work

Summary

Tim O’Reilly argues that the business impact of AI depends less on simply adding AI tools than on reorganizing workflows, data, and incentives around them. Drawing on Trail of Bits CEO Dan Guido’s experience, he describes a progression from AI-assisted to AI-augmented to AI-native work, including a skill-path product that was redesigned as a more interactive experience after the team recognized what language models could already do with contextual information. Adoption is also a human problem: when Trail of Bits began its transformation, 5% of staff were on board, 70% were going through the motions, and 20% were resistant. The company uses a three-level capability ladder, with department-specific matrices, to show employees where they stand. It runs objective-driven hackathons every two months, measures success by movement on the ladder, and stores reusable outcomes in a shared skills repository. To turn failures into infrastructure, it centralizes fixes for prompt debt and uses common sandboxes plus a seven-day cooldown for newly installed external packages. O’Reilly stresses that human expertise remains important, while its Expert MCP server and beta Expert Intelligence initiative aim to ground AI-supported decisions in practitioner knowledge and citations.