Cohere launches its North AI productivity platform into general availability

Artificial intelligence startup Cohere Inc. today made its North productivity platform generally available.

The launch comes about seven months after the offering debuted in early access. Since then, Cohere has built up an initial installed base that includes Dell Technologies Inc., the Royal Bank of Canada and other large enterprises.

North provides a chatbot interface that workers can use to ask questions about their organization’s records. The software is capable of retrieving data from documents, cloud applications and other sources. North can find a file even if a user doesn’t specify its name, but only provides a general description of details such as what information it contains.

The platform’s second core focus is automating content creation. Workers can have North generate financial data visualizations, competitor research reports and other assets. The platform displays a step-by-step overview of how it generates each asset. For example, North might show the SQL query it created to retrieve the data that went into an AI-generated budget proposal.

According to Cohere, companies can also use the platform to automate repetitive business tasks. A sales team, for example, could configure North to sync new  deal details to a customer relationship management system. Cohere says that the platform uses collections of AI agents to automate multistep business chores.

North connects to popular cloud applications using a set of prepackaged integrations. If needed, companies can also link the platform to their custom software via MCP. Originally developed by Anthropic PBC, MCP is a networking protocol that allows agents to take actions in external systems.

Under the hood, North is reportedly powered by a “variant” of Cohere’s flagship Command large language model series. The newest addition to the LLM series, Command A, requires as few as two graphics cards to run. Cohere stated today that on-premises deployments of North can likewise make do with as few as two graphics cards.

Cohere says that Command A can generate prompt responses 75% faster than OpenAI’s GPT-4o model. Furthermore, it mostly outperformed GPT-4o in an internal output quality evaluation. Cohere compared the two models using a set of benchmarks that measure LLMs’ ability to follow instructions, use tools, generate SQL code and power AI agents.

Companies can add guardrails to the AI agents that North uses to automate manual work. Administrators may block agents from performing certain tasks, as well as require them to seek human oversight when carrying out an important activity. In the background, a logging tool tracks service performance and changes to key settings.

The launch of North comes amid reports that Cohere is seeking to raise $500 million in new funding. The investment could reportedly value the company between $5.5 billion and $6.5 billion. Cohere’s annualized annual revenue is expected to top $200 million by year’s end, up from $100 million in May. 

Image: Cohere