Existing Claude Enterprise organizations have until November 16, 2026, to decide whether to accept the default rollout of unified chat and Cowork. Anthropic has made the experience generally available, while its administrator documentation gives organizations options to enable it early or keep it disabled.
The October 23 announcement extends the unified experience introduced for Pro and Max subscribers in September to Enterprise, alongside controls for access, spending and cloud execution. It introduces neither a new Claude model nor the original chat-Cowork merger.
Check Whether Your Organization Will Switch
Anthropic’s Enterprise setup guide says Chat and Cowork unified is off by default in existing organizations and turns on by default on November 16. Owners and Primary Owners can select Enable now or Keep disabled under Organization settings > Capabilities. An existing explicit on-or-off choice stays in place.
Both Chat and Cloud code execution and file creation must be enabled. If either is off, the documentation says nothing automatically turns on. After November 16, enabling both triggers the new experience unless the organization has chosen to keep it disabled.
Start the rollout review with these checks:
- Confirm eligibility. Anthropic says organizations using zero data retention for Claude Code, a HIPAA-ready configuration or customer-managed encryption keys remain on their current configuration.
- Record an activation decision. General availability is not the same as an approved internal deployment.
- Check the prerequisites. The unified experience does not require the older Cowork or Cowork in the cloud switches to be enabled separately.
Organizations that have not enabled unification retain separate chat and Cowork options, as Anthropic’s existing Cowork guidance explains.
One Conversation Can Carry the Project
Employees no longer have to choose between starting a chat and starting a Cowork task. A conversation begins as chat. When the task requires code execution or file work, Claude continues in an Anthropic-managed cloud environment and returns the resulting document, analysis or presentation for review.
Anthropic says that cloud work continues after the user closes their laptop, with the conversation accessible across web, desktop and mobile. Existing chats, projects, connectors, skills and plugins carry over.
Rollout policy needs to address the transition between these activities. Approval to answer questions in a chat interface is not necessarily approval to execute a longer project using connected tools and data, even when both happen in the same conversation.
Assign Access and Budgets Before Broad Deployment
Once the organization-wide switch is enabled, members with the standard User, Admin, Owner and Primary Owner roles receive access. Custom roles need the Chat and Cowork unified capability.
For a limited pilot, administrators should move people who should not receive access into an appropriately restricted custom role before enabling the organization-wide setting. The setup guide warns that new roles start with All capabilities, which includes unification. Creating a custom role alone does not restrict access.
Anthropic lists four usage-management controls:
- Efficient mode keeps the unified experience on Sonnet and Haiku models.
- Model entitlements determine which models each role can use.
- Spend limits apply at organization, group and member levels.
- Pooled group budgets give a group one shared monthly allocation.
Efficient mode constrains model selection without setting a spending ceiling. Model entitlements govern access, while limits and pooled budgets govern expenditure.
Claude Enterprise remains usage-based, and Anthropic cautions that agentic tasks can consume more tokens than quick chats. Administrators should choose model access and budget boundaries together, then observe actual pilot usage. The announcement does not establish a measured savings rate for Efficient mode or an independently verified enterprise productivity gain.
Review the Cloud Boundary, Not Just the Interface
According to Anthropic, each conversation requiring agentic execution receives its own isolated virtual machine. Network egress is off by default for Enterprise, and connector credentials stay outside the sandbox.
These documented architectural protections do not prove that containment cannot fail. Administrators also need to distinguish internet access from connector access: leaving egress disabled does not replace a connector-permission review.
The setup guide’s security settings allow sandbox egress to remain off, permit package managers only, allow a custom domain list or permit all domains. Connector policies are enforced separately, outside the sandbox, on every connector call.
A practical security checklist is to:
