Cohere unveils North 2 AI agent platform with rebuilt orchestration and token spending caps
SiliconANGLE Duncan Riley
Cohere launched North 2, a rebuilt AI agent platform for big companies. It adds tighter spend caps and security controls, which is basically the enterprise AI bill coming due.
Based on reporting by SiliconANGLE, Duncan Riley — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Cohere has rolled out North 2, a major update to its platform for running AI agents inside large organizations. The company is trying to answer a problem that keeps showing up in enterprise AI deployments: pick security, capability or cost control, and you usually end up giving up one of the three. North 2 is meant to make that tradeoff less painful.
The biggest changes are in orchestration and admin control. Cohere rebuilt the system that coordinates multistep agent work, with agents handling tasks on their own and bringing in a human when a step needs sign-off. Administrators also get new token-spending limits in North Admin, the management console. It tracks usage down to individual users and agents, sets tiers for people and groups, caps companywide consumption, and can warn teams before they hit a limit.
There’s also a clear push toward reuse. Agents created from a simple prompt can now be shared across a company instead of rebuilt by different teams. Each agent keeps memory between sessions, and Cohere added a skills feature for reusable logic. Shared libraries hold documents and other assets that every agent should be able to access. North 2 can also turn a natural-language request into a slide deck, dashboard or lightweight app inside the chat window, while templates and a drag-and-drop workflow builder cover recurring work.
The platform is broadening its reach too. The connector library now includes Slack, GitHub, SharePoint and OneDrive, with planned links to financial data providers including PitchBook, S&P Global and FactSet. North remains model-agnostic, so customers can assign models from outside Cohere as well. And Cohere is leaning on Nvidia to reduce running costs, saying its models generate more tokens per second per node on Blackwell and Hopper GPUs.
North still runs on-premises or in a customer’s own virtual private cloud, including fully air-gapped setups. That matters for users like Bell Canada’s Bell Cyber unit, which says the control and sovereignty requirements its customers want are built in from the start. LG CNS uses North internally and for customers in South Korea, while Radical Ventures built an automation that prepares a founder-call briefing. North 2 lands less than three weeks after Cohere signed a merger agreement with Aleph Alpha, with the Cohere name staying on the combined company.
My take — AI-written commentary, not fact-checked reporting
This is the part of enterprise AI that actually matters: not bigger demos, but control knobs that finance teams and security teams can live with. The shiny agent stuff gets the headlines, yet the real product is usually a set of limits, guards and audit trails with a chatbot attached. Boring, yes. Also how these systems get bought.
Read more about this at: SiliconANGLE