Cohere Pivot Toward Enterprise Agents
Cohere at $150M ARR
Cohere is no longer trying to win by being the best general model for every developer, it is trying to win by becoming the software layer enterprises actually deploy. The shift matters because model APIs are increasingly interchangeable, while enterprise buyers pay for private deployment, secure connectors into internal systems, and agents that can search, summarize, and act across company data. That moves Cohere closer to Glean and Writer than to OpenAI and Anthropic.
-
Earlier, Cohere differentiated as a cheaper, faster enterprise friendly LLM provider with private cloud and on premises deployment, while OpenAI and Anthropic were competing more directly for broad model leadership and developer mindshare. That made Cohere a model vendor first.
-
By 2025, Cohere had added North, an agent product that connects to enterprise software and handles longer research tasks. That is the same budget line Glean attacks with enterprise search and copilots, and Writer attacks with enterprise content generation and workflow automation.
-
Cohere’s edge in this new lane is verticalization with owned models. It uses products like Command, Embed, and Rerank, plus design partners like RBC and Fujitsu, to build industry specific systems it can resell across banking or Japanese enterprise without exposing customer data.
The next step is a tighter race between model labs that move up into applications and application companies that move down into owning more of the model stack. If Cohere keeps turning its model work into deployable enterprise agents for regulated industries, it can occupy a defensible middle ground between frontier labs and app only AI startups.