Snowflake Expands Autonomous AI Agent Execution, Strengthens Data and Work Integration
TECHWORLD ·
✦ AI Summary
Snowflake unveiled new features for its enterprise AI agents CoCo and CoWork at Snowflake Expedition 2026.
The company also announced plans to expand Autonomous Data Engineering, Business Ontology, and Cortex AI Gateway, while strengthening data management and governance.
CoCo·CoWork bots will be available in private preview, while CoCo Automation and Slack, Excel, and Google Sheets integrations will be offered in public preview or officially released.
Snowflake is strengthening data management, real-time analytics, and governance capabilities to support autonomous work execution by AI agents. The company envisions AI agents built on enterprise data going beyond repetitive tasks to carry out actual work.
At Snowflake Expedition 2026 on the 4th, Snowflake unveiled new features for its enterprise AI agents CoCo and CoWork. It also disclosed plans to expand Autonomous Data Engineering, Business Ontology, and Cortex AI Gateway.
The focus of the announcement was to support enterprises in delegating work to AI agents while managing data access and execution permissions. Snowflake presented this as part of its Agentic Enterprise strategy.
Snowflake also said it would support data collection, analytics, and work execution on a single platform. Accordingly, it framed strengthened data management, analytics, and governance, along with enterprise control over agent access and execution permissions, as part of its single-platform strategy.
Snowflake added autonomous execution and context-retention features to CoCo, a coding agent for data engineers and developers, and CoWork, a work agent that helps with data analysis and report writing. The goal of the additions is to broaden their use cases.
Snowflake also unveiled "CoCo·CoWork bots." These bots perform background work based on user-defined goals, with examples including preparing materials for customer meetings and checking data pipelines. The feature will be offered in private preview.
Snowflake also officially launched "CoCo Automation," a feature for scheduling repetitive tasks. Data teams can use it to automate pipeline checks and error analysis, while business teams can use CoWork to perform sales performance analysis and write weekly reports.
In this process, the agents also comply with user data access and execution permissions. They are configured to request confirmation when important decisions or final approval are required. In addition, work history is recorded, users can review execution details, and tasks can be stopped.
Snowflake expanded integrations with existing work tools. It made available in public preview the ability to access Snowflake data and run natural-language queries in Slack. It also made available in public preview the ability to access Snowflake data and run natural-language queries in Excel. The same capability was also made available in public preview in Google Sheets.
Snowflake also unveiled a mobile environment that integrates CoCo and CoWork into a single app. As of the end of July this year, there were more than 5,800 active CoWork accounts. As of the end of July this year, there were more than 9,100 active CoCo accounts. These usage figures are based on the average weekly active accounts over the past 4 weeks.
Companies using CoCo and CoWork include Cisco and DocuSign. Christian Kleinerman, Snowflake's senior vice president of product, said accurate data is central to the reliability of AI agents.
Christian Kleinerman, Snowflake's senior vice president of product, said CoCo and CoWork implement agents that understand enterprise work and continue to evolve based on accurate data. Snowflake also strengthened data management and AI governance.
The company also reinforced data processing and operational support features so that AI agents can perform actual work. To that end, it unveiled Autonomous Data Engineering.
In data engineering, agents are responsible for building and operating data pipelines and analyzing errors. If a problem occurs in the pipeline, they analyze the cause and propose fixes. Engineers decide on changes that reflect the actual operating environment. Also, the work environment of the "CoCo Cloud Agent" is configured as a separate cloud that replaces a developer's personal PC, allowing it to continue working even after the user logs off.
The company also expanded real-time data analytics capabilities. The "Snowpipe Streaming" feature collects data generated by applications and devices and scales automatically based on throughput. It also introduced "Snowflake Data Stream," which integrates with existing Apache Kafka-based systems, and the offering will be available in public preview.
The trend is to systematize different business terminology and metric definitions by company to reduce interpretation gaps and analytical inconsistencies among AI agents. To that end, "Business Ontology" is scheduled for official release, with the goal of helping AI agents accurately understand enterprise business rules. This feature manages official definitions of company-specific business terms and metrics, such as revenue and customer classifications, and reduces the likelihood of different analytical results across departments and agents.
Through "Cortex AI Gateway," the company controls agents' access to AI models and also controls external tool calls. In addition, "Cortex AI Gateway" applies enterprise security policies, applies access permissions, manages work history, and manages AI usage costs.
In addition, "Dynamic Model Routing" will be offered in private preview. This feature helps select the right AI model based on accuracy, response speed, and cost. It enables management of operating costs and execution processes alongside the use of multiple AI models.
Snowflake supports workforce development needed as it expands technology adoption through training programs. The company provides Snowflake Practitioner, a free data and AI training program, to customers and partners worldwide.
The program consists of three role-based tracks for data analysts, data engineers, and platform administrators. The training runs for 2 days of foundational instruction and hands-on practice.
CoCo is used throughout the program. The training is offered free of charge through instructor-led sessions and online self-paced learning, and access to the hands-on environment lasts 30 days.
Denise Persson, Snowflake's CMO, said world-class platform education should be the starting point of every customer and partner relationship. She added that Snowflake Practitioner supports the completion of essential introductory training for all customers and partners.
Snowflake is expanding autonomous execution capabilities for AI agents while strengthening its data management and control framework. However, because CoCo·CoWork bots and Business Ontology have not yet been officially released, the company still needs to verify the operational stability of some features in real enterprise environments, and validating the effectiveness of some automation capabilities also remains a challenge.
Source: TECHWORLD · Kim Seung-gi
Original: https://www.epnc.co.kr/news/articleView.html?idxno=409261
References
This article was produced with the help of an automated content generation algorithm.
Source: TECHWORLD
View originalThis article was summarized and organized by BizCrush based on the original article from TECHWORLD. For exact quotations and full details, please refer to the original article.