Cohere Introduces North: An AI Agent Platform Aimed at Protecting Enterprise Data
AI agent tools promise to alleviate some of the tediousness found in daily workflows, yet many organizations remain cautious about their adoption, driven by a significant concern: data security. Large enterprises with trade secrets, firms in heavily regulated sectors, and government agencies have reconsidered the integration of AI tools due to worries that their data—or worse, their customers’ data—might be inadvertently compromised or utilized for training foundational models.
Canadian AI company Cohere aims to address these concerns with its new AI agent platform, North, which assures private deployment, enabling enterprises and governments to safeguard their data along with their customers’ data behind their own firewalls.
“LLMs are only as good as the data they have access to,” Nick Frosst, co-founder and CEO of Cohere, stated during a demonstration of North. “If we want LLMs to be as useful as possible, they must have access to that useful data, which means they need to be deployed in [the customer’s] environment.”
Instead of relying on enterprise cloud platforms such as Azure or AWS, Cohere claims it can install North on an organization’s private infrastructure, ensuring it never interacts with any customer data. North is capable of running on an organization’s on-premise infrastructure, hybrid clouds, VPCs, or air-gapped environments, according to Frosst.
“We can deploy literally on a GPU in a closet that they might have somewhere,” he elaborated, noting that North is designed to operate on as few as two GPUs.
Cohere asserts that North includes security protocols like granular access control, agent autonomy policies, continuous red-teaming, and third-party security audits. Moreover, it complies with international standards such as GDPR, SOC-2, and ISO 27001.
More than private deployments
Cohere, which has secured $970 million in funding with a recent valuation of $5.5 billion, has already piloted North with customers including RBC, Dell, LG, Ensemble Health Partners, and, as reported by TechCrunch last year, Palantir.
North features many capabilities typical of AI agent platforms. Its primary functions include chat and search, enabling users to get answers to customer support queries, summarize meeting transcripts, create marketing copy, and retrieve information from both internal resources and the web. Frosst noted that all responses contain citations and reasoning trails, allowing employees to audit and validate the outputs.
The chat and search functionalities leverage existing Cohere technologies such as Command (its suite of generative AI models) and Compass (its multimodal search technology stack). Frosst mentioned that North is powered by a variant of the Command model tailored for enterprise reasoning.
“It goes beyond just Q&A; it actually performs tasks for you. So, [North] facilitates asset creation. It can generate tables, documents, and slideshows. It can conduct comprehensive market research,” Frosst remarked.
It’s also noteworthy that in May, Cohere acquired Ottogrid, a Vancouver-based platform that focuses on developing enterprise tools for automating advanced market research.
Similar to other AI agent platforms, North can integrate with existing workplace tools such as Gmail, Slack, Salesforce, Outlook, and Linear, and connect with any Model Context Protocol (MCP) servers to interface with industry-specific or proprietary applications.
“As you build confidence through interactions with the model, a smooth transition occurs from using it as an augmentation to employing it as an automation,” Frosst explained.
