Cohere to help businesses build AI using Slack, Google Drive data
Today OpenAI-competitor Cohere, which focuses on large language models (LLMs) for the enterprise, launched “build-your-own connectors” to allow companies to securely connect company data that sits on third-party applications like Slack and Google Drive, Service Now or WordPress to Cohere’s Command LLM.
A company representative said that “as far as we know” the offering is the first of its kind and comes after Cohere recently became “the first and only AI company that offers fine tuning across all four cloud providers.” According to a Cohere blog post, this open-beta release will enable businesses to build AI assistants on Cohere’s upgraded AI platform that leverage information from hundreds of tools used in daily work and provide contextually relevant and accurate responses.
Enables enterprise models to securely access company data
“Data is gold for LLMs,” Roy Eldar, a senior product manager at Cohere who helped lead the project, told VentureBeat. “Enabling enterprise models to securely access company data across dozens of third-party applications is a game changer for what AI can deliver to businesses.”
The blog post said that connectors is “the next step in Cohere’s philosophy of meeting companies where their data is, regardless of which cloud or third-party application they use…allowing companies to plug this data into the models, securely, within their own cloud, and use retrieval-augmented generation (RAG) to see exactly where the answers come from, improving the accuracy of responses, and minimizing hallucinations.”
Cohere explained in the blog post that the upgraded Command framework allows enterprises to connect data from third-party datastores that offer an accompanying search API into RAG. This is an important upgrade because it allows Cohere to display citations next to AI-generated answers so users can click on source links and read original documents for verification or additional context.
Cohere calls connectors ‘dramatic improvement’ for conversational AI
“The seamless integration of RAG in enterprise use cases–and now leveraging data located in third-party data stores via connectors–is a dramatic improvement in the usefulness of conversational AI solutions for businesses,” said the blog post.
To help in-house developers, Cohere said it is releasing about 100 “quick start connectors” on Github for some of the most popular applications where enterprises store data, including GitHub, Asana, Slack, Dropbox, Google Drive and Pinecone. In addition, the company said it can also provide full support for other third-party datastores.
Connectors announcement caps big year for Cohere
The connectors announcements comes at the end of a very big year for the Toronto-based Cohere, which back in February admitted it was flying under the radar before raising a fresh round of $270 million in funding in June and opening up a second headquarters in San Francisco in September.
And after the November drama at OpenAI, during which CEO Sam Altman was fired by the company’s nonprofit board, only to be reinstated five days later, CNBC reported Cohere saying it had “seen increased inquiries from businesses after the barrage of headlines surrounding Sam Altman.” At that time, Cohere co-founder/CEO Aidan Gomez posted a link to the company’s careers page that showed openings for “Machine Learning Members of Technical Staff.”