Cohere took a major step in bringing together enterprise data and AI today with the launch of build-your-own connectors for its Command language model. The new capability allows companies to securely connect proprietary data from third-party business applications to Command, enabling more accurate, contextual answers grounded in real-time company information.
The connectors address a major pain point when deploying AI in business settings. Traditional off-the-shelf natural language models often struggle with queries requiring proprietary data outside their training set. But with custom connectors, Command can now reference data from tools like Slack, Google Drive, Asana, and ServiceNow to provide tailored, relevant responses.
Connectors optimize an AI technique called retrieval-augmented generation (RAG), which Cohere built into Command. Alongside generated responses, RAG provides source citations from linked documents. This reduces "hallucinations", and unveils the "black box" of answer origins common in many LLMs—building user trust through verifiability.
Now with data ingestion from third-party apps, Command can apply RAG seamlessly across proprietary datastores. It's a huge leap in grounding conversational AI to real business needs. Quick answer generation means nothing without confidence in accuracy.
To accelerate development, Cohere is open-sourcing 100+ connector blueprints for leading platforms like GitHub, Slack, Asana, Dropbox, Google Drive, and Pinecone. The templates enable developers to customize connections tailored to company data flows and security protocols. For companies with unique needs, Cohere also offers full support in setting up connectors across various third-party datastores.
The company stresses that the connectors have been designed with rigorous security measures, including user-level permissions that mirror access levels from connected third-party datastores. This approach ensures that sensitive corporate data remains private and secure, addressing one of the major concerns businesses have when deploying AI solutions.
With pre-built integrations now available and customization enabled, connectors promise to rapidly expand Command's capabilities in context-aware enterprise search, decision support, and document understanding. This means more helpful AI assistants that are grounded in organizational knowledge and that can evolve alongside the business.