Snowflake Data Cloud recently unveiled an integration for productivity platform Coda and has now made available for public preview their intelligent SQL (structured query language) queries assistant, Copilot.
Snowflake Copilot first unveiled at Snowday 2017 draws upon Snowflake’s proprietary text-to-SQL model and Mistral’s newly launched Mistral Large LLM to produce relevant SQL queries that help users better comprehend and explore their data. Snowflake recently invested an undisclosed sum in Paris-based startup Mistral Large, investing its entire family of models to its Cortex service for LLM app development.
This decision marks another significant move by the data cloud giant to harness AI to streamline how enterprises utilize their data assets – an initiative spearheaded by Sridhar Ramaswamy after taking over as CEO from Neeva AI acquisition.
What Can Snowflake Copilot Offer Enterprise Users? Snowflake stands at the forefront of the data revolution, helping enterprises analyze assets and extract meaningful insights for decision making. Their unique approach has propelled their growth; however, extracting insights has often required writing complex SQL code, which takes time and may not suit every enterprise user. With Snowflake Copilot they aim to make extraction simpler with less writing involved while still meeting enterprise user requirements for insights extraction.
Snowflake is meeting this challenge head on with their newly introduced Copilot, now rolling out in public preview for AWS accounts in the U.S. The assistant sits silently within SQL worksheets and offers users a conversational interface for creating SQL queries in natural language – simply pull up Copilot via “Ask Copilot” button and state what you need in English; soon afterwards it understands your question, processes it, and returns a ready-to-use SQL code ready to run in order to achieve desired results.
Snowflake reports that customers can leverage Copilot’s generation capabilities for various tasks, including extracting data from multiple tables for analysis as well as correcting existing queries. If users don’t know where to start, Copilot provides assistance by engaging in dialogue to understand the structure of a dataset and what questions need to be asked to gain insights. It also understands context by responding with detailed responses about which tables were joined before providing answers.
Snowflake utilized Cortex, its own service for LLM app development, as well as vast amounts of SQL query data and metadata from Mistral Large for this endeavor.
“With over 4 billion queries running daily on our platform, Snowflake AI provides unprecedented insight into even the most complex data challenges. This massive amount of data fuels Copilot development beyond typical large language models,” according to Pieter Verhoeven and Yusuf Ozuysal from Snowflake AI respectively in their joint blog post.
Snowflake plans on expanding Copilot through public preview. Snowflake intends to use user feedback from this trial version to fine-tune and ready its solution for general availability; although its exact timeline remains unknow. For now, Copilot remains focused on SQL worksheets; however, Snowflake has hinted that at some point soon, Copilot could expand to cover other aspects of their product and become an “ubiquitous companion” of sorts for users.
Though details regarding what areas Snowflake will expand into are yet to be decided, its result can easily be anticipated: ease of accessing insights from Snowflake for faster decision-making. While this could be game-changing for the company, Snowflake is not alone when it comes to natural language querying technology – LLMs have also allowed many data players to explore simplifying product experiences with generative AI, while Dremio and Kinetica both recently introduced similar capabilities for querying using LLMs.
