The useful question is what changes for users, developers or buyers, and whether the announcement stays industry context or becomes something people can actually use.

Whatnot, a rapidly expanding live-shopping marketplace, joined Snowflake at Summit 2026 to share how it transformed its data operations to maintain visibility and reliability amid explosive growth. The company processes over 550,000 hours of livestreams weekly and once hosted a U.S. record-breaking stream with 583,000 concurrent viewers, generating billions of daily data points that power its live app experience. Engineering Manager Alice Leach emphasized that delays in data processing directly impact customer interactions, requiring real-time insights rather than historical reports.

Originally centralized under a single data team using dbt, Whatnot’s infrastructure became a bottleneck as the company scaled. To address this, it adopted a modular data stack, allowing business units like fraud prevention and vendor analytics to independently spin up dedicated Snowflake warehouses on demand. While this decentralization improved agility, it introduced new challenges in maintaining visibility, cost control, and performance across the organization, particularly as data scientists became overwhelmed with ad hoc requests.

Whatnot’s solution evolved through three phases: an AI Slackbot in 2024, decentralized tools like Sigma and Glean in 2025, and in 2026, Hex Threads—a custom AI companion powered by Snowflake Cortex Agents. The agentic analytics tool scans the data network to act as a conversational assistant, eliminating the need for users to know SQL or debug errors. Within 90 days, over 80% of Whatnot’s 1,000+ employees adopted the tool, with 17 departments achieving 100% utilization, enabling self-sufficiency in areas like marketing and talent acquisition.

To manage decentralized AI adoption, Whatnot implemented a monitoring system designed for speed, cost efficiency, and clarity. Standard logs were delayed by hours, while frequent diagnostic scans were compute-intensive and required admin access. The company now balances rapid data access with operational control, embedding Snowflake Cortex Agents with row-level security in its Seller Hub to allow livestreamers to query data instantly while protecting privacy and managing costs.