Once seen purely as a data warehousing powerhouse, Snowflake is undergoing a major reinvention. At its Snowflake Summit 2025, underway in San Francisco, the company unveiled a sweeping set of AI products that reimagine how data is ingested, processed, and turned into intelligence, all within one unified platform.
Traditional ETL (Extract, Transform, and Load) processes often involve integrating multiple separate tools, such as Talend, Informatica for data integration, Airflow for orchestration, and Spark for processing, to build complex data pipelines. These tools are typically combined to handle extraction, transformation, and loading tasks, which can lead to complexity, higher costs, and maintenance overhead.
On the other hand, Snowflake’s Openflow, a new multimodal ingestion service powered by Apache NiFi, helps enterprises pull in data from diverse sources and formats into Snowflake’s AI Data Cloud.
“This is the productisation of an acquisition we made a few months ago of a company called Datavolo. Openflow is a managed service that helps organisations both extract data from a variety of sources and be able to process it,” said Christian Kleinerman, EVP of product at Snowflake, in a media briefing.
Openflow allows customers to move data from where it is created to where it is needed, supporting both batch and streaming modes. It features hundreds of pre-built connectors and processors, and offers extensibility to build custom connectors. The service supports Snowflake’s Bring Your Own Cloud deployment model and is now generally available on AWS.
Moreover, the platform removes the existing bottlenecks in data engineering, including rigid pipelines, fragmented stacks, and slow ingestion. Openflow supports both structured and unstructured data and integrates with sources like Box, Google Ads, Oracle, Salesforce Data Cloud, Workday, and Microsoft SharePoint.
“Most of our customers are interested in loading data into Snowflake or making it available to Snowflake,” said Kleinerman. He further added that their goal is to simplify data movement and processing from any one source to any other destination.
With Openflow, Snowflake is also extending its data engineering capabilities. Customers will soon be able to run dbt Projects natively in Snowflake with support for features like in-line AI code assistance and Git integration.
The capability will be available within Snowflake Workspaces, a new file-based development environment. These projects will eventually be powered by dbt Fusion.
Snowflake also announced expanded support for Apache Iceberg tables, which allows organisations to build a connected lakehouse view and access semi-structured data using Snowflake’s engine. New optimisations for file size and partitions are expected to improve performance and control.
Snowpipe Streaming, now in public preview, adds support for high-throughput, low-latency data ingest, with data becoming queryable within 5 to 10 seconds. This further improves Openflow’s ability to manage near-real-time data streams.
From Data to Action
Besides, Snowflake has announced new agentic AI offerings at its annual user conference, including two innovations called Snowflake Intelligence and Data Science Agent.
Snowflake Intelligence, launching soon in public preview, allows non-technical users to query and act on structured and unstructured data through natural language prompts.
The product is powered by Cortex Agents and LLMs from OpenAI and Anthropic, and runs directly inside customers’ Snowflake environments, inheriting security and governance controls.
“Snowflake Intelligence breaks down these barriers by democratising the ability to extract meaningful intelligence from an organisation’s entire enterprise data estate — structured and unstructured data alike,” said Baris Gultekin, head of AI at Snowflake.
Snowflake Intelligence also incorporates third-party content through Cortex Knowledge Extensions, including CB Insights, Packt, Stack Overflow, The Associated Press, and USA TODAY.
On the other hand, Data Science Agent automates core machine learning tasks using Claude from Anthropic. These tasks include data preparation, feature engineering, and model training. The agent provides verified ML pipeline code and allows users to iterate through suggestions or follow-ups.
“We’re leveraging AI to help customers create machine learning pipelines, writing code, validating it, and ultimately automating the end-to-end ML lifecycle,” said Kleinerman.
The company claims the agent reduces the time spent on debugging and experimentation, allowing data scientists to prioritise higher-impact work.
These launches are part of Snowflake’s broader push to enable enterprise AI use cases. For analytics, Snowflake has also launched AISQL, which extends its SQL language to include AI operations as simple function calls.
“The goal of this is to bring the power of AI to analysts and personas that are typically comfortable with database technology,” Kleinerman explained. This includes processing text for sentiment analysis and classification, and supporting multimodal data like PDFs, audio, and images.
Analysts can now enrich tables with chat transcripts, correlate sensor data with images, and merge structured data with sources like social media sentiment—all in one interface.
The tool integrates with sources like Box, Google Drive, Workday, and Zendesk using Snowflake Openflow and supports natural language conversations that return insights, generate visualisations, and surface business knowledge.
The company also introduced SnowConvert AI, an agent that automates data migrations from platforms such as Oracle, Teradata, and Google BigQuery. It reduces the need for manual code rewriting and validation, and accelerates database, BI, and ETL migration processes by two to three times.
“SnowConvert AI enables organisations to quickly and easily move from legacy data warehouses… while staying supported and without disrupting critical workflows,” the company said.
With these launches, Snowflake is moving beyond the traditional data warehouse, positioning itself as a full-stack AI platform for enterprises, spanning ingestion, processing, and intelligent automation.
The post Snowflake Just Killed the Data Pipeline as We Know It appeared first on Analytics India Magazine.
