Cohere, a Canadian AI firm, has launched North, a new AI agent platform designed to alleviate data security concerns hindering AI adoption in large enterprises and highly regulated industries.
The primary concern among organizations has been the potential for inadvertent data compromise or the use of proprietary or customer data to train foundational models. North addresses this by offering private deployment, allowing enterprises to keep their and their customers’ data securely behind their firewalls. Nick Frosst, co-founder of Cohere, stated, “LLMs are only as good as the data they have access to. If we want LLMs to be as useful as possible, they have to access that useful data, and that means they need to be deployed in [the customer’s] environment.” Unlike solutions relying on public enterprise cloud platforms, Cohere asserts that North can be installed directly on an organization’s private infrastructure, ensuring Cohere never accesses or interacts with customer data. North supports various deployment environments, including on-premise infrastructure, hybrid clouds, Virtual Private Clouds (VPCs), and air-gapped environments. Frosst highlighted North’s efficiency, noting it was engineered to operate on as few as two GPUs, stating, “We can deploy literally on a GPU in a closet that they might have somewhere.”
To bolster its security claims, Cohere states that North incorporates robust security protocols, including granular access control, agent autonomy policies, continuous red-teaming, and third-party security tests. North is designed to comply with stringent international compliance standards, such as GDPR, SOC-2, and ISO 27001. Cohere has raised $970 million and was recently valued at $5.5 billion, and has initiated pilot programs for North with prominent customers, including RBC, Dell, LG, Ensemble Health Partners, and Palantir.
Functionally, North offers features common to many AI agent platforms, with primary capabilities centered around chat and search. These features enable users to handle customer support inquiries, summarize meeting transcripts, generate marketing copy, and access information from internal resources and the web. A key transparency feature is that all responses generated by North include citations and “reasoning” chains of thought, empowering employees to audit and verify output for accuracy and accountability. The chat and search functionalities are powered by Cohere’s proprietary technologies: Command, its family of generative AI models, and Compass, its multimodal search tech stack. Frosst clarified that North utilizes a specialized variant of the Command model, trained for enterprise reasoning.
Beyond standard Q&A, North is designed for asset creation, generating tables, documents, and slideshows, and performing market research, reinforced by Cohere’s acquisition of Ottogrid in May. North offers extensive integration capabilities with common workplace tools, connecting seamlessly with applications like Gmail, Slack, Salesforce, Outlook, and Linear. It can also integrate with Model Context Protocol (MCP) servers, accessing industry-specific or in-house applications. Frosst noted the progression in user interaction, observing, “As you build confidence by chatting to the model, there’s like a smooth transition that happens between using this as an augmentation to using it as an automation.”
