The cloud data company is introducing autonomous engineering workflows and real-time analytics capabilities, addressing the growing challenge of keeping enterprise information accurate and available for AI systems.
Cloud data platform Snowflake has introduced autonomous data engineering capabilities that allow artificial intelligence agents to monitor data pipelines, identify failures and propose repairs without requiring engineers to supervise every step.
The November 3 announcement was made during Snowflake Expedition 2026, where the company unveiled enhancements to its CoCo coding agent alongside new real-time analytics and data integration features.
The initiative represents another development in enterprise AI infrastructure: moving from tools that help programmers write code toward systems that can independently maintain recurring technical operations.
For technology companies, the question is whether these capabilities can reduce infrastructure maintenance while preserving reliability, security and engineering control.
AI Agents Take On Recurring Engineering Tasks
Snowflake’s CoCo Cloud Agents operate in isolated cloud environments rather than requiring continuous execution on individual developers’ computers.
The agents can access approved project files, investigate pipeline problems and work within restrictions established by data engineers.
Through CoCo Automations, teams can schedule recurring tasks such as checking pipeline health or investigating failed jobs.
According to Snowflake’s technical documentation, the system can identify likely causes of failures and suggest code changes for engineers to review before deployment.
Integration with existing development workflows, including GitHub, is intended to preserve approval processes.
Snowflake executive vice president of product Christian Kleinerman argued that increasing data volumes and AI adoption are making manual pipeline maintenance increasingly difficult to sustain.
Real-Time Data Becomes a Competitive Requirement
The announcement also introduced capabilities designed to support applications requiring continuously updated information.
Snowflake said its real-time analytics infrastructure can support ingestion rates of up to 20 gigabytes per second per table, with a median data-readiness latency of three seconds.
Those performance figures come from Snowflake’s internal testing and should not be interpreted as independently verified results across all workloads.
The company also expanded support for Apache Spark and Apache Iceberg, aiming to reduce the code changes required when moving existing data-processing workloads onto Snowflake.
Customer Cboe Global Markets reported bringing nearly 200 billion rows of market data into Snowflake daily, while financial technology company auxmoney described using Pipeline Builder to simplify its data engineering processes.
Security Remains a Critical Limitation
Autonomous maintenance creates additional risks when AI agents receive permissions to inspect files, modify code or interact with enterprise systems.
In an October interview with TechRadar, Snowflake chief security and trust officer Mayank Upadhyay emphasized the importance of controlling agent permissions and maintaining human oversight.
He identified both external attacks and unintended actions by an organization’s own agents as security concerns.
Several newly announced Snowflake capabilities remain in public or private preview, although CoCo Cloud Agents and CoCo Automations are generally available.
AI Is Becoming Part of Core Software Infrastructure
Snowflake’s expansion follows strong commercial growth associated with enterprise AI adoption.
In September financial reporting, Reuters documented a 37% increase in Snowflake’s second-quarter product revenue and the company’s decision to raise its fiscal 2027 product-revenue forecast to $6.07 billion.
Chief executive Sridhar Ramaswamy attributed approximately half of the acceleration in growth to AI offerings.
For engineering leaders, the new products introduce another decision: whether autonomous maintenance can reduce operational costs without weakening review procedures or increasing dependence on a single platform.
The broader significance is that enterprise AI competition is expanding from generating software toward maintaining the infrastructure that software and AI agents depend upon.
