Full Breakdown
Cohere Unveils North 2, an Agent-Centric AI Platform for Enterprises
By Drooid · · How we work
North 2 Platform Unveiled
Cohere Inc., the Toronto-based AI firm, has launched North 2, the second generation of its North platform. The upgrade shifts the system from a primarily generative chatbot that could retrieve internal documents to an “agentic” service capable of executing multistep tasks such as resolving customer issues, performing financial analysis, and writing code. The company says the new version already ships to customers and is built around an “agent harness” that links AI models to external applications.
Background and Market Context
Founded in 2019, Cohere targets regulated industries, public-sector clients, and large enterprises that prioritize privacy and control. Its first North release arrived last year. The firm competes with U.S. rivals OpenAI and Anthropic, which offer enterprise-focused products like ChatGPT Enterprise and OpenAI’s Dots. Cohere is expanding its reach through a merger with German AI company Aleph Alpha, giving it a foothold in the European market.
Key Features and Security Controls
North 2’s agent harness manages autonomous agents and enforces security policies. Customers can restrict which documents and tools agents may access, require human approval for high-risk actions, and set cost-monitoring limits to curb expensive, inefficient task execution. The platform is model-agnostic, allowing integration with any large language model, including those from competing providers.
Customer Adoption and Revenue Model
Cohere lists Bell Canada, LG CNS Co. Ltd., and Dell Inc. as North users, though it has not disclosed the total client count. The company reports “dozens of companies” with some deployments reaching “tens of thousands of users.” A sizable portion of Cohere’s revenue derives from the harness that orchestrates agents rather than from model licensing alone.
Implications for Enterprise AI
By emphasizing agentic behavior, built-in security, and cost controls, Cohere aims to differentiate itself in a market where AI agents have drawn scrutiny for misbehavior and high operational expenses. If the harness model proves effective, it could influence how enterprises balance autonomous AI capabilities with governance requirements.
Verbatim Quotes
- “The model is like the engine, and the harness is like the vehicle,” — Ms. Pineau
- “We have dozens of companies, and some of them are up to tens of thousands of users,” — Ms. Pineau
- “We actually derive a sizable portion of our revenue from the harness,” — Ms. Pineau