Cohere views enterprise search as utility
Product manager at Cohere on enterprise AI search infrastructure and deep research agents
This shows that AI search APIs are becoming utility layers, not strategic battlegrounds, for enterprise model companies like Cohere. North’s main job is helping large companies search their own internal systems, reason over that information, and deliver useful work products in secure environments. In that context, paying Tavily for web retrieval is easier than staffing a separate crawling and search team, because vendor swap pain is low and the real product differentiation sits higher up the stack, in connectors, orchestration, deployment, and enterprise workflow fit.
-
Cohere already evaluated providers, moved from Brave to Tavily, and found the main benefit was easier integration. Tavily returned usable page text for LLM grounding, while Brave required extra fetching and transformation work. That makes the choice operational, not existential.
-
The same interview makes clear that customer demand is centered on internal document grounding, not best in class public web search. Fortune 500 deployments live inside customer environments, often with customer owned API keys, so web retrieval cost stays modest while the hard work is making SharePoint, internal files, and other proprietary systems usable for agents.
-
This is consistent with the broader market. Exa sells meaning based web search as a core product and argues retrieval is where most of the value sits. But even there, search providers look interchangeable enough that buyers like Cohere and Ecosia keep architectures flexible. The durable value for enterprise vendors is the full workflow around search, not search alone.
Going forward, the likely winners will bundle retrieval into bigger enterprise systems rather than try to win on raw search alone. Cohere is moving in that direction with North, while search specialists like Exa, Parallel, and Tavily will keep supplying pieces of the stack until domain specific data access and agent orchestration become important enough to justify deeper vertical integration.