Preparing for AI Agent Era with Trustworthy Data and Governance

CoCo, CoWork, and Cortex AI Gateway Unveiled

"AI strategy can ultimately only be completed on top of a data strategy"

Global AI data cloud company Snowflake held its annual conference "Snowflake World Tour Seoul 2026" on the 27th at COEX in Samseong-dong, Seoul.

On the keynote stage, with "Agentic Enterprise" as the new corporate operating model, Lee Yong-seok, Korea Country Manager of Snowflake, Bala Kasiviswanathan, Vice President of AI and Developer Experience, and Lee Soo-hyun, Tech Evangelist, took the stage in turn to introduce Snowflake's vision and actual implementation cases.

Lee Yong-seok, the Country Manager who took the stage first, introduced the growth trajectory of the Korean market.

He said, "Since the launch of the Korea branch in 2021, domestic sales have grown approximately 18-fold by 2025, and the number of customers has increased more than 10-fold," and added, "Eight of Korea's top 10 conglomerates as well as major digital companies such as Nexon, Kakao, and Toss are actually operating data and AI workloads."

He pointed out that AI adoption is still in the proof-of-concept (PoC) stage at many enterprises.

As data is moved externally and separate environments are built, security verification and operational complexity increase, weakening implementation momentum.

In response, he said, "Snowflake will enable AI to be built at its original location without moving data, supporting rapid transition from experimentation to actual operating environments."

Passing the microphone to Vice President Bala Kasiviswanathan, he defined the future of enterprise AI as "Agentic Enterprise."

He said, "Today, enterprises find themselves in complex environments with numerous applications and databases mixed with different data formats," and "For AI to operate safely and reliably, simply implementing a model is not enough."

Snowflake presented four solutions.

First is trustworthy data and context, second is flexible AI model selection not locked to a specific vendor, third is connectivity with enterprise business applications, and fourth is a control plane that integrates and manages all AI agents.

Snowflake presented a strategy to provide these four elements in an integrated manner on top of its AI data cloud.

Particularly noteworthy was the AI agent governance framework.

Snowflake recently acquired Natoma, a specialist in AI agent identity and access control, and integrated it with its own platform.

Additionally, through "Cortex AI Gateway," the company enabled centralized control of access rights, costs, and activity logs of various AI models, applications, and agents within the enterprise.

Following the theoretical explanation, as the stage darkened, the audience's attention turned to the large screen.

Tech Evangelist Lee Soo-hyun demonstrated an Agentic Enterprise scenario using a hypothetical sports and entertainment company as an example.

In the demo, a scene was shown where integrated analysis was performed on a single screen combining sales data, operational information, CCTV footage, and fan application review data to analyze the causes of sales increases.

AI analyzed real-time streaming data, unstructured video data, and customer response data simultaneously to derive insights, and subsequently results were automatically transmitted via Slack and email.

In another scene, the newly launched AI agent for developers "CoCo" appeared.

When a user requested in natural language to "protect the data," the AI automatically identified sensitive information and generated and applied appropriate masking policies. Subsequently, a process was demonstrated where the concept of "sales" used by different departments with different standards was defined through a semantic view, enabling AI to understand the same business language.

The audience was particularly focused on scenes where AI performed actual actions beyond simple analysis.

Vice President Bala emphasized, "An era is opening where every employee can create their own applications and work automation," and "An era is arriving where even non-developers can directly build data-driven services."