How would you measure the success of Cohere North?

Introduction

North is Cohere's agentic AI workspace for enterprises. Employees use it to chat with company knowledge, search documents through Compass, generate reports, and run agents that take actions across tools like email, CRM and ticketing systems. It runs on Cohere's own models (Command A, Embed and Rerank) and can be deployed in the customer's private cloud or on-premises. Cohere made North generally available in August 2025, and it has since become central to the company's story: Cohere is no longer just a model API company but a provider of a full enterprise AI platform.

Measuring North's success is tricky. The buyer (a CIO or business unit leader) and the user (an employee) value different things, and the product runs in environments where Cohere may not see detailed usage. I'll define success with a clear framework that works for both.

Step 1

Product Context (3 minutes)

Key stakeholders:

• Employees (end users): want faster answers and less busywork.
• Business sponsors (CIO, COO, line-of-business heads): want measurable productivity gains and a return on AI spend.
• Customer IT and security teams: want control, auditability and no data leakage.
• Cohere: wants seat expansion, renewals and reference customers.

User flow:

1. IT deploys North and connects data sources (SharePoint, Google Drive, Salesforce, Outlook and others).
2. Employees log in through single sign-on and see only what they are permitted to see.
3. They ask questions, search with Compass, or start agents and workflows.
4. North returns answers with citations or completes multi-step tasks for review.
5. Admins monitor usage, cost and audit logs.

Product lifecycle stage: Early growth. North has marquee customers in banking, telecom and the public sector, and Cohere's partnership with Bell Canada brings it to Canadian organisations through a sovereign AI offering. The priority now is proving value within accounts and expanding from pilots to company-wide rollouts.

Competitive context: Microsoft 365 Copilot, Google Gemini Enterprise, ChatGPT Enterprise and Glean. North differentiates on private deployment, model ownership and retrieval quality.

Software-specific context:

• Deployment: customer VPC, Cohere-managed isolated environments, or on-premises.
• Telemetry: often aggregated, with customers controlling what is logged. Metrics must be computable without reading prompts.