Technology · NEWS EXPLAINER
Cohere unveils North 2 with new controls for workplace AI agents
The October 5 announcement emphasizes orchestration and usage controls. Earlier workflow features and planned connections need separate treatment.
Cohere announced North 2 on October 5, describing a redesigned system for coordinating multistep AI work. The enterprise platform's administrative controls include user quotas, request and token rate limits, and organization-wide caps. The announcement concerns how workplace agents operate and are managed; it is not a new consumer chatbot plan.
Context and control serve different jobs
Cohere describes memory that retains context between sessions, libraries for shared knowledge and assets, and skills for reusable capabilities. North Admin separately provides visibility into users, agents and token consumption, with thresholds for usage alerts.
The distinction matters when evaluating a workflow: an agent's ability to remember context does not determine what it may access or how much it may consume. Those are separate questions for the people configuring it.
July's announcement already included automations
Cohere's July 27 post said automations were available to North customers. It described scheduled workflows, model selection at individual steps, approval gates and monitoring of input and output tokens. October 5 should therefore not be presented as the first appearance of every workflow feature.
The older post also described planning and testing a workflow before putting it into production. A useful rollout discussion can ask what a proposed workflow will do, where a person must approve it, and how an administrator will inspect its use. The dated posts do not establish the first shipment date of every individual North 2 capability.
Planned connectors are a separate category
The October announcement lists workplace connections including Slack, SharePoint and GitHub. It describes financial-data providers such as PitchBook and FactSet as planned additions, without a launch date.
For a team depending on a particular data source, being named in a roadmap is a reason to ask about availability. It is not sufficient evidence that the connection can be used in a deployment today.
Limits answer a different question from prices
A cap or rate limit constrains consumption; a price sets what that consumption costs. The controls described in this announcement do not by themselves demonstrate a price increase, a discount or a guaranteed final bill.
Our earlier reports on OpenAI Dots and NVIDIA's agent safety platform examine related questions about approvals and enforcement. They provide background on workplace agent controls, rather than evidence that those products share North's features.
