Should Cohere prioritise Managed North or customer-hosted private deployments for its next platform investment?

Introduction

Cohere sells North in two broad ways. In customer-hosted deployments, North and Cohere's models run inside the customer's own cloud account or data centre, managed by the customer's platform team with Cohere's support. In Cohere-managed deployments (often described as Managed North), Cohere runs a dedicated, isolated environment for the customer, so the customer gets the same privacy guarantees without operating the infrastructure.

Both matter. Customer-hosted deployments are the core of Cohere's appeal to banks, governments and defence agencies. Managed deployments are faster to launch, easier to upgrade and better suited to mid-size companies. The prompt: "Cohere has engineering capacity for one major platform investment next year. Should it prioritise Managed North or customer-hosted deployments?"

Step 1

Clarifying Questions (3 minutes)

• Context: What is today's revenue and pipeline mix between the two models?

Why it matters: Shows where the business stands and is heading. Expected answer: About 80% of current revenue comes from customer-hosted or private deployments, but 55% of new pipeline opportunities by count are mid-size companies asking for a managed option. Impact on approach: Current revenue favours customer-hosted; future growth may favour managed.

• Business: How do margins and time to revenue differ?

Why it matters: Profitability matters as Cohere prepares for a possible public listing. Expected answer: Customer-hosted deals have higher contract values but take about 7 weeks to reach production and require heavy solutions-engineering support. Managed deals go live in about 2 weeks but Cohere carries GPU costs. Impact on approach: We need to compare contribution margin, not just revenue.

• User: What do customers say they need?

Why it matters: Avoid building for the wrong buyer. Expected answer: Regulated enterprises demand control; mid-size customers want "private, but someone else runs it." Impact on approach: Segment our analysis.

• Technical: How much shared work exists between the two?

Why it matters: Some investments (packaging, upgrades, observability) help both. Expected answer: About 40% of planned work would benefit both. Impact on approach: The real trade-off is the remaining 60%.