Snowflake's EVP of Product Talks Hard Truths on Agentic AI: Readiness, Governance, and AI Economics

Most agent demos look impressive. Far fewer survive contact with messy data, unclear permissions, brittle workflows, and unpredictable costs. Christian Kleinerman, EVP of Product at Snowflake, lays out the hard truths: where autonomous workflows are producing measurable value, where the industry is overselling, and why many initiatives fail for reasons unrelated to model quality.

This conversation focuses on the practical requirements for deploying AI agents at scale.

The conversation covers:

Data readiness: Why agents fail without a unified strategy for structured and unstructured dataGovernance as an accelerator: Closing the trust gap so agents can move from answering questions to taking actionAI economics: Moving from prepaid commitments to measuring ROI and controlling the cost of agentic workflows in productionWorkforce impact: How agentic automation reshapes roles, team design, and accountability when agents absorb work once done by teams

Join this live conversation and direct your questions to Christian for real answers. Take advantage of this opportunity!

Key TakeawaysData Quality Remains the Single Biggest Barrier to AI Success

Organizations fail to extract value from AI agents when their data estates lack organization, governance, and a single source of truth.

Kleinerman states directly that if data is siloed, for example, with inconsistent customer lists or unclear ownership, no AI model will produce reliable results. Leaders must prioritize data rationalization, establish canonical data sources, and enforce clear security policies before expecting AI initiatives to succeed. 

The most common conversation Snowflake has with customers centers on accelerating data quality efforts to make their data "AI-ready." Treat data preparation not as a preliminary task but as the foundation upon which all AI value depends.

Replace Lengthy Proofs-of-Concept with Rapid Iteration and Low-Cost Experimentation

The economics of evaluating AI technologies have shifted dramatically. What once required months of planning and dedicated hardware now takes an hour of hands-on testing. 

Kleinerman advises leaders to try three or four technologies in a day or two rather than running traditional three-month proof-of-concept cycles. This approach reveals genuine capability differences obscured by vendor benchmarks and marketing claims. 

The friction of evaluation has dropped significantly in the AI era, making direct empirical testing the only reliable method for assessing fit.

Most AI Projects Fail When Moving from Pilot to Production

The transition from proof of concept to production exposes critical gaps in data maturity and system complexity. 

Pilots may work with five well-named tables, while production environments may contain hundreds of thousands of cryptically labeled data assets. Kleinerman identifies two consistent failure modes: incorrect results eroding user trust and security violations exposing information to unauthorized users.

Organizations that built their own solutions 12 to 18 months ago now face harsh reality checks during production rollouts. Leaders should anticipate this friction and plan for the messy, scaled conditions of real enterprise environments from the outset.

Episode Participants

Christian Kleinerman serves as Snowflake’s EVP of Product and has been with the company since 2018. He oversees the company’s global product strategy and vision. Christian is a database expert with over 20 years of experience working with various database technologies and has more than 15 years of management and leadership experience. Most recently, Christian worked at Google leading YouTube’s infrastructure and data systems. Prior to that, he served as General Manager of the Data Warehousing product unit at Microsoft where he was responsible for a broad portfolio of products. Christian holds a BS in Industrial Engineering from Los Andes University in Colombia, and he is a named inventor on numerous Snowflake patents.

Michael Krigsman is a globally recognized analyst, strategic advisor, and industry commentator known for his deep business transformation, innovation, and leadership expertise. He has presented at industry events worldwide and written extensively on the reasons for IT failures. His work has been referenced in the media over 1,000 times and in more than 50 books and journal articles; his commentary on technology trends and business strategy reaches a global audience.