Tim O'Reilly interviews Box CEO Aaron Levie at AI Codecon, exploring the paradox of rising software engineering job postings despite AI disruption fears. Levie argues AI agents will expand demand for engineers across all enterprise functions via a Jevons paradox effect, not shrink it. Key themes include the real bottleneck being enterprise data context (not connectivity), the trillion-dollar question of when to use deterministic vs. probabilistic AI systems, and why startups will win in unstructured workflow domains like legal and accounting. Levie also highlights that agents lack the ambient human context that workers carry for free, requiring enterprises to fundamentally reengineer workflows rather than bolt AI onto existing processes.