Enterprise agents stop trading off: North 2 puts reusable skills, shared libraries and per-agent guardrails in one platform

Enterprises deploying agents no longer have to choose between security, intelligence, cost and control, because North 2 places them in one platform. The prior choice was that a platform secure enough was not smart enough, one smart enough was not controllable enough, and one governable became too expensive to run at scale. Cohere states that over the past year it rolled out North across finance, healthcare, telecommunications, manufacturing, energy and the public sector, investing in large-scale production deployments that shaped North 2.

The platform breaks agent capability into reusable parts, so employees do not rebuild the same logic repeatedly. Previously similar workflows often had to be rebuilt, and agents started cold every session. North 2 offers six core feature groups: Skills, Libraries, Memory, Applications, Automations and Connectors; Memory keeps context across sessions, and Connectors cover Slack, SharePoint, OneDrive, Microsoft Outlook, Microsoft Exchange, Jira, Linear, Notion and GitHub, with planned financial data providers including PitchBook, Crunchbase, Daloopa, FiscalAI, S&P Global and FactSet.

Security controls reach the individual agent level, so an enterprise can roll agents out to a workforce without rolling out risk alongside them. Agents that touch internal tools and models exposed to prompt injection changed the risk calculus for enterprise security teams. North deploys privately on-prem, in the cloud or fully air-gapped; North Admin enforces guardrails at the individual agent level; the platform holds SOC 2 Type 2, ISO 27001 and ISO 42001 certifications and uses continual security testing, red-teaming and third-party vulnerability scanning.

Admins can see and control AI usage down to the user and agent level, including token spend and limits. Previously it was hard to map AI investment to specific users, agents or models. North Admin sets roles and permissions, controls which models are used where and by whom, provides real-time activity analytics and defines token consumption tiers; Cohere states that Cohere models trained and deployed on NVIDIA Blackwell and Hopper increase tokens per second per node, decreasing spend.

The report covers Cohere's release of the North 2 enterprise agent platform, its feature list, certifications and customer collaborations. It is intended for researchers and engineers evaluating enterprise agent deployment. The material states no next experiment or validation plan.

The effectiveness statements come from Cohere and its collaborations with NVIDIA, LG CNS and Bell Cyber, without reproducible benchmarks or quantitative comparisons; the NVIDIA-related statement says North generates more insights for fewer tokens but does not disclose the measurement conditions.