Did you know AI agents not only talk to you, but they go and perform tasks for you! Stanford's newest human-centered AI report reveals exciting advancements in agentic AI, showing how these smart agents are able to explore, plan, and learn on their own. In fact, the report reveals autonomous agents can already handle some real-world tasks like research and online shopping.

SaaS businesses are at the forefront, already using smarter, self-directed AI to streamline workflows by combining data from various sources. For example, AI agents could power solutions to automatically create tasks, update records, or trigger notifications based on specific events or conditions. The possibilities are absolutely inspiring.

With models, like our family of Command R models, that excel in data retrieval and tool use, businesses can empower their AI solutions with autonomous features and actions using retrieval-augmented generation (RAG) systems.

Start Building AI agents with RAG

Keep it simple.

Creating reliable and high performing AI agents can be challenging work. Start with a simple execution. RAG naturally extends to richer executions across multiple data sources, enabling AI agents to pull in diverse insights and provide more comprehensive solutions. With multi-step tool use, large language models can leverage external tools or APIs in order to handle a variety of tasks beyond text generation. Here’s a simple multi-step notebook to get you started, or watch Cohere’s Jay Alammar explain the basics in this video.

Build faster.

Focus on task-specific agents with a defined set of actions that an AI agent can take given its state and environment. A deterministic approach can provide guardrails for agentic behaviors, ensuring more predictable and controlled actions. We’ve recently launched the Cohere Toolkit to help you build advanced AI solutions faster. It’s a repository of production-ready applications deployable across cloud platforms, and it comes with all the elements to create an AI assistant, the first step toward building agentic capabilities.

Consider the holistic impact.

A technology capable of planning and executing sequences of actions for individuals necessitates careful consideration of the wider societal, workforce, and personal impacts. Features like more autonomy, the ability to chat in natural language, and high levels of customization can make AI agents incredibly useful. However, these same features could also leave people open to undue influence by the technology, so it's important to think about introducing safeguards right from the start. To better understand this, and many more technical governance topics in AI, check out the recently launched series from Cohere For AI where they explore the landscape of AI policy and governance.

It is still early days for autonomous workflows, where flexible agents are capable of continuous learning and have the capacity for long-term reasoning and decision-making, but the promise it holds for Enterprise AI is clear. BCG’s report on the coming agentic revolution says it best, “Autonomous agents can directly ‘tell’ other enterprise systems what to do. This could fundamentally change how a company operates, enabling it to deploy automation more holistically.”

Product

This month, we launched our most advanced generative model, Command R+, on the Microsoft Azure cloud platform. The Command R model family is also available on Amazon Bedrock. And we added a new, faster, and better Rerank 3 model for improved retrieval with RAG.

For Business

Read our new series of industry explainers showcasing how individual sectors are using GenAI to improve productivity and drive innovation. Starting with the insurance industry, legal services industry, and SaaS sector we explore real-world use cases.

Developers

Join the private beta for Cohere Compass, our new foundation embedding model that allows indexing and searching on multi-aspect data.

Research

Cohere For AI released the weights for Command R+ to be used for research purposes. The model has over 150K downloads already! Try it yourself.

Company

Cohere achieved the Amazon Web Services (AWS) Generative AI Competency in the categories of Consulting Services, Generative AI Applications, Foundational Models and App Development, and Infrastructure and Data.

AI agents are coming