Cohere Inc. today introduced North, a new product designed to make it easier for business workers to use its artificial intelligence models.

The company describes the offering as a “secure AI workplace platform.” It’s available through an early access program.

Toronto-based Cohere develops a series of large language models called Command. Unlike OpenAI, the company doesn’t offer a consumer chatbot service but rather focuses solely on the enterprise market. Cohere received a $5.5 billion valuation last July following a $500 million funding round.

North, the company’s new product, provides a chatbot interface that allows customers to interact with its Command LLM series. Workers can use the tool to analyze earnings reports, find documents and perform other business tasks. North can include not only text in its prompt responses but also other data such as graphs.

Users can customize North by creating AI agents, AI programs optimized for a narrow set of tasks. A human resources professional, for example, could build an agent that automates some of the work involved in onboarding new hires. Cohere says that creating an AI agent takes a few clicks and doesn’t require programming expertise.

In an internal evaluation, the company compared North with Microsoft Copilot and Google Vertex AI Agent Builder across a set of finance, HR, customer support and IT tasks. Cohere says that its product outperformed the competition across all four categories. The test used a benchmark called Llama Index that accepts an AI-generated answer if it’s “at least relevant and correct.”

Under the hood, North is powered by not only Command but also Cohere’s Compass product. The latter offering is an AI-powered search tool that can find specific data snippets in a company’s systems and make them available to AI models. It extracts data from not only documents but also multimodal files such as slides, images and spreadsheets.

Compass is based on two AI models. The first, Embed, turns data into embeddings, mathematical representations that AI models can process more easily than raw business documents. The other model, Rerank, analyzes the information that is retrieved in response to a user query and prioritizes the most relevant items. 

According to Cohere, North can be deployed both in the cloud and on-premises. It’s capable of running in air-gapped environments, environments that are isolated from the rest of a company’s network and the web for cybersecurity reasons. That could make it easier for organizations in highly regulated industries to adopt North. 

“The current build-it-yourself approach to AI deployment places a huge burden on organizations to invest the time, expertise, and resources into developing bespoke solutions and then maintaining them,” Cohere founder and Chief Executive Officer Aidan Gomez wrote in a blog post. “North helps avoid these pain points and reduces the time to value for customers.”

Cohere says that the Royal Bank of Canada is one of the early customers using North. According to the LLM provider, it will work with the bank to develop a version of the software optimized for the financial sector. That raises the possibility Cohere could potentially launch additional industry -specific versions down the road.

The introduction of North comes about two months after OpenAI debuted canvas, a productivity-focused interface for ChatGPT. The feature is geared towards writing and programming tasks. It includes a large interface panel dedicated to displaying the user’s code or text, as well as AI-generated editing suggestions.

Image: Cohere