Connectors Are Key to Enterprise Search
Product manager at Cohere on enterprise AI search infrastructure and deep research agents
This points to connectors being part of Cohere’s core product, not just a nice add on. Cohere has already bought Ottogrid and is folding its workflow, data extraction, and research tooling into North, while also positioning North and Compass around built in connectors for enterprise systems. The bigger point is that enterprise retrieval is won less by generic search quality and more by handling messy, customer specific systems like SharePoint correctly inside secure deployments.
-
The acquisition most likely refers to Ottogrid, which Cohere bought in May 2025. Ottogrid automated research workflows, extracted structured data from websites, and enriched records, then Cohere said the product would be incorporated directly into North.
-
The interview makes clear why this is hard to commoditize. A connector is not just a login to SharePoint or another app. It has to pull the right files, parse inconsistent formats, respect permissions, and deal with every customer having a different internal setup.
-
That puts Exa and Parallel in a mixed role. Better connectors could be useful infrastructure if they are cheaper or better, but they do not fully displace Cohere when the customer wants private deployment, custom implementation, and end to end enterprise search over internal systems.
Going forward, connectors will become a bigger battleground inside enterprise AI assistants. The likely winners are the vendors that combine secure deployment, strong permissioned retrieval, and practical workflow tools that turn scattered internal data into answers, reports, and actions without forcing enterprises to rebuild their stack.