AI agent tools promise to streamline daily workflows by reducing repetitive tasks, yet many organizations remain hesitant to adopt them due to concerns about data security. Large enterprises, companies in regulated industries, and government agencies worry that sensitive data—especially customer information—could be compromised or used to train foundational AI models.
Canadian AI firm Cohere aims to address these concerns with its new AI agent platform, North, which enables private deployment. This allows enterprises and governments to keep their data secure behind their own firewalls. Cohere co-founder and CEO Nick Frosst emphasized that large language models (LLMs) are only as effective as the data they access, making deployment within a customer’s environment essential for maximizing utility.
Unlike relying on public cloud platforms like Azure or AWS, Cohere can install North directly on an organization’s private infrastructure, ensuring the company never interacts with or stores customer data. North is designed to run on various setups, including on-premise infrastructure, hybrid clouds, virtual private clouds (VPCs), and even air-gapped environments. Frosst noted that the platform can operate on as few as two GPUs, even in minimal setups like a single GPU in a closet.
Security is a key focus for North, featuring granular access controls, agent autonomy policies, continuous red-teaming, and third-party security testing. The platform also complies with international standards such as GDPR, SOC-2, and ISO 27001.
Cohere, which has raised $970 million and holds a $5.5 billion valuation, has already piloted North with customers like RBC, Dell, LG, Ensemble Health Partners, and Palantir.
North offers standard AI agent functionalities, including chat and search capabilities that help users answer customer support queries, summarize meetings, draft marketing content, and retrieve information from internal and external sources. Responses include citations and reasoning chains, allowing employees to verify outputs.
The platform leverages Cohere’s existing technologies, such as Command (its generative AI model family) and Compass (its multimodal search stack). A specialized variant of Command powers North, optimized for enterprise reasoning. Beyond basic Q&A, North can create assets like tables, documents, slideshows, and perform market research—a capability enhanced by Cohere’s recent acquisition of Ottogrid, a market research automation platform.
Like competing AI agent platforms, North integrates with workplace tools such as Gmail, Slack, Salesforce, Outlook, and Linear, and connects to Model Context Protocol (MCP) servers for industry-specific applications. Frosst described a seamless transition from using the platform for task augmentation to full automation as users build confidence in its capabilities.