Cohere has updated its North 2 platform to function as a central hub for enterprise AI agents, capable of managing complex, multi-step workflows while preserving context across separate sessions. The revamped system coordinates various agents that access shared knowledge libraries and reusable skills, linking them to external tools such as Slack, SharePoint, and Jira. Users can generate presentations, dashboards, and simple applications directly through text prompts within the chat interface. The software supports deployment on-premises, in the cloud, or in fully air-gapped environments to meet specific security requirements.
The update emphasises model-agnosticism, allowing organisations to integrate Cohere’s Command A+ models alongside their own proprietary systems. Administrators utilise the North Admin interface to control token usage, user quotas, and access rights for individual agents. Critical actions initiated by agents require human approval before execution. Cohere states its models require fewer tokens when running on Nvidia’s Blackwell and Hopper hardware. Current production customers include LG CNS and Bell Cyber. The company targets government agencies and regulated industries seeking strict control over AI operations, a strategy aligned with its recent acquisition of Aleph Alpha.
- Deployment options include on-premises, cloud, and air-gapped environments
- Critical agent actions require mandatory human approval
- Token efficiency improves on Nvidia Blackwell and Hopper hardware
