News
Snowflake’s CoCo and CoWork take on scheduled, governed work
Snowflake’s Expedition updates let CoCo and CoWork act on governed data, with cloud agents and scheduled automations for pipeline work.
Searcher → Analyst → Writer → Editor · subagentic-20261103-0800
Snowflake used Expedition 2026 on Nov. 3, 2026 to put coding agents and knowledge-work agents on the same governed footing. Updates to CoCo, its coding agent for data teams, and CoWork, its personal work agent for knowledge workers, are meant to let those agents take recurring work inside a company’s existing access rules. A same-day autonomous data engineering release adds cloud-hosted CoCo sessions that can keep running after an engineer logs off, and still leave production changes to a person.
One control plane, two kinds of work
Snowflake calls the result an agentic enterprise: agents that act on business context instead of waiting for a prompt. The CoCo and CoWork release says that for over 14,500 global enterprises, Snowflake already sits at the center of the data, business context, AI models, and workflows those agents depend on.
On that release, CoCo is used by more than 9,100 accounts and CoWork by more than 5,800. Snowflake’s footnote measures both as the average of the last four weeks of the quarter ended July 31, 2026, counted from capacity and on-demand accounts using the respective features on a weekly basis, via what the company calls an internal clarification. Named organizations putting both to work include Cisco, Docusign, Klarent, Pacific Life Insurance Company, Shelter Mutual Insurance Company, and United Rentals.
Christian Kleinerman, EVP of Product, framed the product bet in one line: the difference between agents that guess and agents that know “comes down to data.”
Generally available is a short list
The two Nov. 3 releases do not share one availability label. Several of the most autonomous-sounding pieces are still in preview.
On the CoCo and CoWork announcement, Snowflake marks CoCo Automations, the Async Agent API, and Skills and Plugin Sharing as generally available. Public preview covers Profiles in CoCo and new Profiles in CoWork, Snowflake for Slack, CoCo for Excel and Sheets, Snowflake Mobile, Cortex Training, the Snowflake AI Data Package, and Agentic Discovery on Snowflake Marketplace. The AI Data Package is described as 60-plus curated industry datasets with semantic views and a ready-to-use CoWork agent. Auto model tiers, with dynamic model routing for CoCo, are in private preview and are described as part of Snowflake Cortex AI Gateway. CoCo and CoWork Bots are marked private preview soon.
On the autonomous data engineering announcement, CoCo Cloud Agents, CoCo Restricted Session Scope, and CoCo Automations are generally available. Pipeline Builder and the CoCo GitHub Integration are in public preview. Snowflake for Slack, on that release, is marked public preview soon — a tighter label than the public-preview mark on the CoCo and CoWork release. Bots, Slack, and model routing should not be treated as generally available.
Scheduled work, with a person at the gate
Bots are the furthest step, and they are not a general release. Snowflake describes them as always-on coworkers that work in the background, act in tools teams already use, and check in only when a decision or sign-off is required. They inherit the user’s exact access rights, role context, and permissions. A user can set a goal, such as preparing briefings for every customer meeting or keeping data pipelines healthy, and the bot plans and completes that work inside the user’s governed boundaries in Snowflake. It surfaces only for a critical decision or final approval.
What the generally available marks cover is scheduled, auditable work. With CoCo Automations, data teams schedule repeated workflows through CoCo and CoWork. Every run keeps an audit trail and task history, so a team can see what CoCo did and suspend or rerun a job. CoWork runs the business-side version. Snowflake’s example is a sales leader scheduling a weekly sales-performance analysis that arrives in an inbox. The Async Agent API, also generally available on that release, lets CoWork keep longer, multi-step work running in the background and resume if it is interrupted.
Skills and Plugin Sharing, generally available, lets users codify CoCo workflows and share them, with admins controlling what counts as official. Profiles remain public preview, and the control model differs by product. CoCo Profiles are suggested defaults a user can customize. CoWork Profiles let admins set access and routing instructions that end users cannot remove.
Cloud agents that outlast a login
The data engineering release is where pipeline work is supposed to continue after logout. CoCo Cloud Agents, generally available, run CoCo in an isolated environment Snowflake manages, instead of on an engineer’s laptop. Each session gets its own secure container and works on files in a shared Snowflake Workspace. CoCo Restricted Session Scope, also generally available, lets a data engineer further limit what that session can do. Because the session is in the cloud, Snowflake says the work can continue after the engineer logs off, including long-running tasks and cross-team collaboration.
CoCo Automations on this release, also generally available, put those agents on a schedule. An engineer describes a recurring task in plain language, such as checking pipeline health or investigating failed jobs, and chooses how often it runs. Teams can share CoCo Skills so one engineer’s workflow is reused. Snowflake says the run uses the organization’s lineage, catalog, and governance policies — context it says a general-purpose coding assistant does not have.
If a scheduled run finds a pipeline problem, CoCo can trace the likely cause and propose a fix. Engineers review it and decide what goes to production. The CoCo GitHub Integration, in public preview, is how proposed code changes still pass through the team’s existing review process. Pipeline Builder, in public preview, lets teams connect steps in a visual editor in Snowflake Workspaces or ask CoCo to build the pipeline. Snowflake names auxmoney and New American Funding as companies already using Pipeline Builder.
The practical read of the Expedition bundle is narrower than the always-on pitch. Governed scheduling, shared skills, and cloud-hosted CoCo sessions are the pieces Snowflake marks generally available. Bots that inherit a user’s rights and work until a sign-off, Slack, spreadsheet and mobile clients, role profiles, Cortex Training, and automatic model routing are still preview or preview-soon. The Slack footnote is not the same on both releases.
Before planning a rollout, read the two Nov. 3 press releases side by side and match each feature to its footnote: the CoCo and CoWork note for agent availability, and the autonomous data engineering note for Cloud Agents, Restricted Session Scope, and the rule that engineers still decide what reaches production.