Excerpts from the Interview with the CIO of Snowflake
What changes have you noted since your appointment as CIO at Snowflake over the past six months?
The CIO role at Snowflake is a departure from traditional norms, particularly due to my software-focused background. Previously, as CIO of payments technology at J.P. Morgan, I developed software solutions tailored for their payments clients. This deep engagement in building customer-facing software has shaped a solutions-oriented mindset.
At Snowflake, I have sought to extend that mindset internally. Our approach treats all company departments and functions as our customers. We foster active relationships with them, aiming to understand their goals and objectives. This allows us to drive innovations that address their present needs while also preparing for future challenges.
What are your main objectives for the upcoming year?
One significant goal is to streamline our legacy applications. Currently, we operate around 365 applications, which is excessive for a workforce of 8,500 employees. I believe we should aim for a more manageable number, closer to 150. We are committed to consolidating thoughtfully, ensuring we consider future needs and maintain an AI-first perspective.
Additionally, we are focused on scaling agentic AI. At present, we have approximately 35 agents functioning at varied levels of maturity across different business units, including sales, legal, finance, and HR. In the next year, we plan to expand this to between 50 to 75 agents. These agents are designed not just as productivity tools; they aim to transform employee interaction with data and systems. Our development will continue to emphasize enhancing the accuracy of data exchange.
How do you envision the future of data interactions with agentic AI?
We anticipate seeing the emergence of micro agents, which will involve breaking down existing agents into smaller, more specialized units that can operate with higher precision. By assembling several specialized micro agents—each potentially achieving 98% accuracy—we could enhance overall accuracy compared to a generalized agent that might only reach 80%. This necessitates a rules-based engine to coordinate the agents effectively.
Overall, the evolution of data interactions hinges on striking a balance between access, accuracy, and user experience. For CIOs, the task extends beyond simply developing agents that provide answers; it also encompasses building systems that instill trust among employees.
What is the primary challenge regarding agentic AI that you are currently addressing?
A significant challenge lies in determining how an AI agent should respond when it cannot provide an answer due to inaccuracies in data or restrictions on the user’s access rights. This issue transcends basic data permissions—it encompasses the privileges linked to specific answers. We are dedicating substantial resources to resolving this internally while piloting some of our initial agents.
We handle sensitive data related to our customers and their clients, including medical information, where privacy rights and access controls are paramount. Our focus includes critical considerations such as how to ascertain the inquirer’s identity, their rights to specific data, and how to respond appropriately when answers cannot be provided. Simply issuing an ‘error’ message is inadequate; we must handle these situations strategically.
We must ensure that users receive appropriate responses without compromising access restrictions, all while preventing frustration that could steer them back to traditional dashboards.
As access to data becomes easier, what new risks does this present?
The simplification of data access through AI has unveiled data risks more rapidly and on a larger scale. Previously, an employee needed SQL skills to query sensitive data; today, they can simply interact with a chatbot in everyday language to access restricted sales data. While issues of poor governance and inconsistent access rights are not new, they have become more conspicuous with the advent of AI.
The challenge for CIOs is to simplify data access while maintaining strict oversight around privileges, privacy, and adherence to corporate and regulatory standards. Consequently, reinforcing role-based access and compliance has become increasingly critical.
