Industry practitioners and experts discuss the shift from viewing shadow AI as a security threat to recognizing it as a critical demand signal. The conversation moves beyond blanket bans toward building secure, high-speed ecosystems that empower innovation while maintaining rigorous governance.

The speakers

• Manish Agarwal: Global VP of Enterprise Data Analytics, bringing a perspective from data & analytics organization enabling fast, trusted, and scalable decision-making
• Bijit Ghosh: Former Managing Director, CIO – Technology leader in the Financial Services sector, focused on baked-in governance and platform-based approaches for regulated industries.

Moderated by Rahul Parwani, Head of Product at Airia.

Key discussion pillars:

1. Redefining shadow AI

Shadow AI occurs when the sanctioned organizational tools fail to meet user needs in terms of speed, capability, or accessibility. Rather than just a policy failure, leaders should view it as a mismatch between demand and supply.

2. Governance as an enabler

The panel advocates for a shift from a “blocker” mindset to making a “safe yes” possible. Key strategies include:

• Invisible governance: Building architecture (like SSO and platform-level guardrails) so that security is seamless for the end user.
• The hub-and-spoke Model: Centralizing a governed platform (the hub) while allowing individual business lines to tailor workflows (the spokes).
• Adaptive architecture: Ensuring systems are model-agnostic and swappable to keep up with the rapid pace of AI evolution and model deprecation.

3. Managing the AI lifecycle

AI governance is a continuous lifecycle, not a one-time audit.

• Start small: Begin with low-risk, high-value “pilot” use cases to prove value and build internal trust.
• Technical guardrails: Protecting data from being used in model training and implementing a strong semantic layer to maintain data accuracy and access controls.
• Day two operations: Preparing for the challenges of “change management,” such as when models update and change their verbosity, cost, or behavior.

Core insight for leaders

“Shadow AI isn’t primarily a security problem, it’s a demand signal… [it tells you] that the sanctioned options aren’t meeting [users] where they are.” — Rahul Parwani

“AI is shifting from insights to decisions, from dashboards to conversations and automation. As it begins to generate context and summaries, governance ensures this intelligence is trusted, controlled, and scalable.” — Manish Agarwal

The path forward for CDOs is to close the trust and speed gaps by educating users, building robust data boundaries, and providing a “safe path” that makes rogue AI unnecessary.