The rise of AI agents creates a tension in system design, requiring frictionless access while maintaining robust controls and limitations. Authentication methods for AI agents face unique challenges, including accessing multiple systems programmatically, maintaining persistent access without human intervention, and securing credentials and other secrets. The traditional boundaries between human and programmatic access are blurring as advanced AI models interact with systems through user interfaces rather than APIs alone. Organizations must rethink authentication paradigms to balance security with operational effectiveness, considering best practices such as managed identity services, robust credential rotation, and minimal privilege models.