Toronto-based Cohere released North 2 on Monday, Oct. 5, an upgraded version of its enterprise AI platform that lets companies build and run AI agents for multi-step work while controlling security and costs.
Cohere called the release its biggest upgrade yet in a blog post. It said the platform combines enterprise-grade security, cost governance and full-stack control, and can run wherever a customer’s data lives.
According to MobileSyrup, North 2 lets agents work independently on complex processes, share reusable agents and knowledge assets across an organization, and use session memory so automations improve over time. It also adds a drag-and-drop workflow builder and real-time performance monitoring.
The platform is model-agnostic, meaning customers can use Cohere’s own models or those from competitors. Paul Teyssier, Cohere’s vice-president of product for AI, said clients will keep using Cohere’s technology for most regular AI applications but could choose alternatives for experimentation and niche tasks, MobileSyrup reported.
On security, MobileSyrup reported that a North Admin tool lets organizations monitor AI usage, and that guardrails can be set for individual agents to prevent data leaks. The platform also logs agent activity and applies autonomy policies that limit agents to authorized actions. Cohere said it has done red-teaming exercises and third-party vulnerability scanning.
Bell Cyber is using North 2. MobileSyrup reported that Bell Cyber partnered with Cohere to bring government-grade security to agentic AI. Pricing and general availability dates were not disclosed in the coverage reviewed by Canada News Media.
North first reached the market in limited form in January 2025. Cohere widened its release on Aug. 6, 2025, when the Canadian Press reported it was already deployed at Bell, the Royal Bank of Canada and Dell.
Co-founder Nick Frosst described North at the time as a tool for automating work people would rather not do behind a computer. He said Cohere could pull from a customer’s data without moving it out of the customer’s environment, so the company does not see it.
That data-residency pitch is central to Cohere’s Canadian strategy. Banks, telecoms and public-sector bodies handle regulated information, and many want AI tools that keep that data inside their own systems rather than in a foreign cloud.
The launch follows Cohere’s partnership with OpenText on agentic AI for regulated sectors, which Canada News Media reported on Sept. 22. It also comes as Bell builds out AI infrastructure in Saskatchewan, with Cohere as one of its partners.
The model-agnostic design is a notable shift for a company that built its reputation on its own language models. It suggests Cohere is betting that the control layer around AI agents, not only the models underneath, will be what enterprise customers pay for.
Independent confirmation of customer results was not available. The details of the agent features come from Cohere’s announcement and one news report, so the product’s real-world performance remains to be seen.
If you find reporting like this useful, consider supporting Canada News Media, an independent Canadian newsroom with no corporate parent.
Via MobileSyrup (original report), with information from Cohere and the Canadian Press.
