Part 7/12:
To make this vision viable, security, permissions, and governance will be paramount. Nadella discusses trust boundaries—users and organizations will need control over what AI agents can access and do with sensitive data. Microsoft is already exploring this with tools that govern permissions in enterprise AI applications, ensuring users retain control over their data and AI actions.
The Broader Industry Transition
Nadella references that Microsoft's AI business is already a significant revenue stream—around $10 billion—focused mainly on inference workloads via APIs rather than raw training capacities. This suggests a shift from building large models from scratch to utilizing foundational models and building intelligent agents on top.