Cohere is an artificial intelligence (AI) company that develops large language models and related tools for businesses and organizations. Its technology can help computers understand language, generate written content, search large collections of information, summarize documents, and answer questions using an organization’s own data. This article was written by ChatGPT.
Although Cohere is sometimes compared with companies such as OpenAI, Google, Anthropic, and Microsoft, it has placed particular emphasis on enterprise AI. Rather than focusing mainly on a public chatbot, Cohere provides technology that organizations can integrate into their own websites, software, internal systems, and business processes.
Where Did Cohere Come From?
Cohere was founded in Toronto in 2019 by Aidan Gomez, Nick Frosst, and Ivan Zhang. The founders had experience working with artificial intelligence and machine learning, including work connected to Google Brain.
Aidan Gomez was also one of the authors of the influential 2017 research paper Attention Is All You Need. That paper introduced the Transformer architecture, which became an important foundation for many modern large language models.
Transformers made it possible for AI systems to analyze relationships between words and pieces of information more effectively. This technology helped support the development of today’s generative AI systems, including tools that can write, summarize, translate, classify, and answer questions.
Cohere grew out of this period of rapid progress in language-based artificial intelligence. Its founders saw an opportunity to make advanced language models available to organizations through practical software tools and application programming interfaces, commonly called APIs.
What Does Cohere Do?
Cohere develops AI models that businesses and developers can connect to their own applications. These models can perform several kinds of language-related tasks.
For example, an organization might use Cohere to:
- Create an internal assistant that answers employee questions
- Search reports, policies, manuals, or research documents
- Summarize long pieces of text
- Classify emails, support requests, or customer feedback
- Generate drafts, descriptions, and other written material
- Help users find the most relevant information in a large database
The organization using the technology still needs to decide what information the AI can access, what tasks it should perform, and how its answers will be reviewed.
Cohere’s Main Types of AI Tools
Cohere has developed several categories of tools for working with language.
Command Models
Command models are designed to understand instructions and generate responses. They can support conversational assistants, question-answering systems, summarization tools, and other generative AI applications.
For example, a company could use a Command model to build an assistant that answers questions about its workplace policies.
Embed Models
Embed models convert words, sentences, and documents into numerical representations called embeddings. These representations help computers identify similarities in meaning.
This makes it possible to search by concept rather than relying only on exact keywords. A search for “help paying rent,” for example, might also locate a document about housing assistance even when the exact wording is different.
Rerank Models
Rerank models help improve search results. After a search system finds a group of possible documents, a reranking model evaluates them and moves the most relevant results toward the top.
This can be especially useful when an organization has thousands of documents and needs to find the best information for a particular question.
Using an Organization’s Own Information
One of Cohere’s most important business applications is helping organizations use AI with their own documents and data.
A general AI model may know a great deal about public information, but it will not automatically know the contents of a company’s internal reports, training manuals, project records, or policies.
Organizations can connect an AI system to approved information sources so it can retrieve relevant material before generating an answer. This approach is commonly known as retrieval-augmented generation, or RAG.
For example, an employee might ask:
“What is our policy for booking vacation time?”
The system could search the organization’s current human-resources documents, locate the relevant policy, and use that information to prepare an answer.
This approach can make an AI assistant more useful because its responses are based on information that is specific to the organization.
Why Is Cohere Considered Enterprise AI?
Enterprise AI refers to artificial intelligence designed for use within businesses, governments, educational institutions, nonprofits, and other organizations.
Enterprise systems often need features that are less important in a casual public chatbot. These can include:
- Access controls
- Security and privacy protections
- Integration with existing software
- Support for large document collections
- Reliable search and retrieval
- Administrative oversight
- Options for where and how the technology is deployed
Cohere has positioned its products around these organizational requirements. Its technology may operate behind the scenes inside another company’s application, so a person could use a service powered by Cohere without seeing the Cohere name.
How Is Cohere Different From ChatGPT?
ChatGPT is a public-facing AI assistant that people can use directly for writing, research, brainstorming, coding, learning, and many other tasks.
Cohere is more commonly presented as a collection of AI models and tools that developers and organizations use to build their own systems.
The two companies therefore overlap in some areas, but their products are often experienced differently. A person can open ChatGPT and begin a conversation. With Cohere, the AI may be integrated into an employer’s document-search system, customer-service platform, or internal assistant.
This distinction is not absolute. AI companies continue to expand their products, and their features may change over time. However, Cohere’s enterprise focus remains one of the clearest ways to understand its place in the AI industry.
Could Smaller Organizations Use Cohere?
Cohere’s technology is not limited to very large corporations. Smaller businesses, nonprofits, libraries, municipalities, educational institutions, and social enterprises could also use enterprise AI.
A community organization might create a searchable assistant for its programs and services. A library could improve searches across local resources. A municipality could help staff locate policies and planning documents. A nonprofit could organize research, grant information, and internal reports.
The main challenge is not simply gaining access to an AI model. An organization also needs suitable information, technical support, clear privacy rules, human oversight, and a well-defined purpose.
Why Cohere Matters
Cohere represents an important part of the development of artificial intelligence: the movement from general-purpose AI demonstrations toward tools that can be incorporated into everyday organizational work.
Its history also connects directly to the Transformer technology that helped make modern generative AI possible. From those research beginnings, Cohere developed into a company focused on helping organizations search, understand, and work with language at scale.
For beginners, the main point to remember is this: Cohere develops language-based AI tools that businesses and organizations can use to build their own search systems, assistants, and information services.
In the next article in this series, we will compare Cohere and ChatGPT and look at how their typical uses differ.