Making internal knowledge the product
Product manager at Cohere on enterprise AI search infrastructure and deep research agents
Cohere is competing less on having smarter models in the abstract, and more on doing the messy enterprise plumbing that turns a company’s own files into reliable answers. In practice that means connecting North into systems like SharePoint and other proprietary stores, pulling the actual document text, adapting to each customer’s custom setup, and tuning retrieval so a bank, healthcare company, or government team can trust what comes back from its own internal corpus.
-
The hard part is not generic web search. It is enterprise specific retrieval. Cohere already uses an outside vendor for web search, and the product team says the main effort is making internal document grounding work across customer by customer implementations, where even common systems like SharePoint are configured differently at every large company.
-
That focus lines up with where Cohere is headed as a business. North is positioned as an enterprise assistant for knowledge workers, and Cohere’s broader shift has been toward private cloud and on premises deployments for regulated customers that want their data to stay inside their own environment.
-
It also explains the competitive split with Glean and Writer. Glean’s wedge is broad internal search across SaaS apps, Writer’s wedge is compliant content generation, while Cohere is trying to own more of the full stack, model, retrieval, deployment, and vertical tuning, so internal knowledge quality becomes the product, not just a feature.
The next step is that enterprise AI platforms will be judged less by demo quality and more by how well they can read a company’s real systems without leaking data or missing key context. As North expands from search into longer running agents, the winner will be the vendor that makes internal knowledge feel complete, current, and trustworthy enough to automate real work.