What is most important for securely scaling AI-driven automation?

Ric Smith

President of Products and Technology · Okta

Identity

Ric Smith says identity is foundational for securely scaling AI-driven automation, because identity has always been the last perimeter—a basic principle he says enterprises forget when focusing on technical abilities.

Rama Akkiraju

VP, Enterprise AI · Nvidia

Observability

Rama Akkiraju says observability is key, especially for understanding agent behavior before an agent can be certified.

JA

James J Azar

Chief Information Security Officer · Artera Services

Understanding Risk

James J Azar says understanding risk may be the most important point from his CISO perspective, because security's responsibility is to establish controls proportionate to the risk created by business decisions about AI.

Mandy Andress

CISO · Elastic

Contextual Data

Mandy Andress says the answer is contextual data: the more context you provide about your own environment and organization, the less risk of AI hallucinations and wrong conclusions.

The answer starts at 29:54.

Vipin Samar

Senior Vice President, Database Security · Oracle

Secure Data at Source

Vipin Samar argues the only real defense is to secure data at source—putting security in the database where data belongs—because agentic AI lacks discipline and will bypass application-level controls.

Julian Tang

COO for Innovation Office · BlackRock

Unified Infrastructure

Julian Tang says enterprises can only scale AI when fragmented data, legacy systems and siloed ownership are addressed as a single infrastructure problem rather than as isolated technical fixes.

Jacob Thomas

Manager of Information Security · Texas Children'S Hospital

Governance Model

Jacob Thomas says a good governance model behind everything matters, citing mitigating and compensating controls that, if planned correctly, reduce the attack surface and threat actors' ability.

Joshua Aaron

CEO · Aiden Technologies, Inc.

Human Oversight

Joshua Aaron says human oversight is needed—a human in the loop—because full AI making these decisions might rush to fix something and break more than it fixes, so his team has humans decide and automation execute quickly.