Connect AI Agents to Your Enterprise with Sema4.ai Actions

Introducing Sema4.ai Actions

Two years ago, when ChatGPT first emerged, the AI community faced a crucial challenge: enabling language models to programmatically interact with enterprise systems. Today, as AI agents become increasingly sophisticated, developers need robust tools to bridge the gap between AI capabilities and enterprise infrastructure.

Enter Sema4.ai Actions – a comprehensive framework designed for developers building AI agents that need to interact with enterprise systems at scale.

This blog post explores the three key features of the Sema4.ai Actions one by one:

• Action Gallery: Pre-built enterprise integrations to common tools like Microsoft Office, Google Workspace, HubSpot and much more.
• Automation-as-code platform: Leverage the power of Python tailored for agent use-cases. Our platform makes it a breeze with dependency management, secrets, logging all handled for you.
• Custom action SDK and VS Code extension: Build your own actions to connect agents with your applications in any way you want. Whether it’s SAP, Workday, or your proprietary software, if your business uses it, your agents can too.

Action Gallery

Sema4.ai Actions provides an extensive collection of pre-built actions that enable seamless integration with common enterprise apps like Microsoft SharePoint, Office, Google Workspace, HubSpot, and more. 

These ready-to-use actions allow you to connect your AI agents to essential business platforms effortlessly, without writing a line of code. We provide all the actions as open-source code through the Gallery GitHub repository, providing the developers an option to fork and extend any of the pre-built actions – a unique feature not available in any other AI Agent framework.

Automation-as-code platform

Our platform offers a full-featured automation-as-code solution, utilizing the extensive capabilities of Python. With an integrated enterprise Python environment management system, Sema4.ai Actions handles managing and isolating dependencies, securely managing secrets, and providing comprehensive logging and debugging.

So what does it really do for the developer? Here are my personal favorites:

• Managing the entire Python environment and all dependencies in one package.yaml file (using the ultrafast uv instead of pip) completely eliminates the mess with virtual environments. Our tooling ensures that actions built on dev’s laptops run the same way in the production, completely solving the “…well, it worked on my machine…” Python problem.
• Every action run produces a detailed execution log, without me needing to clutter my code with print statements.
• Built-in secrets management that allows me to easily declare the API keys, secrets, and OAuth2 clients that actions need to connect to enterprise applications. Locally, I can have my dev keys, and the Sema4.ai platform handles the secrets securely all the way to production environments.
• Automatically generated OpenAPI spec from docstrings. AI Agent is using these actions based on the descriptions given to it in the “tool” specification. Our platform generates these descriptions automatically from the docstrings and type hints in the code. No additional boilerplate code required.

Custom action SDK and VS Code extension

Sema4.ai Actions offers an extensive SDK and Microsoft VS Code extension. The actions really come alive in VS Code – our recommended IDE to build custom actions tailored for complex enterprise environments and apps like SAP, ServiceNow, Workday, and more.

VS Code Extension handles the entire lifecycle of developing actions with built-in features for:

• Bootstrapping projects from powerful templates
• Real-time code validation that keeps you on track and reminds if you miss any critical element
• Integrated run/debug functionality and access to action execution logs without leaving VS Code
• Management of test data inputs for the actions
• Publishing actions upstream to your organization

But what actually makes Python code an AI Action? It’s really simple, actually. Having the project structure in place, for example from our templates, it’s a matter of adding a simple @action decorator, and you are done! Like this:

We believe that AI agents’ true potential lies in their ability to work alongside existing enterprise systems, and Sema4.ai Actions makes this integration not only possible but simple and secure.

To illustrate the power of Actions, let’s consider a real-world example. One of our customers, a large manufacturing firm, used a combination of Actions for Microsoft tools from the Gallery with custom Actions that wrap their existing custom AWS Lambda functions that interact with their ticketing system. The benefits of Sema4.ai Actions were immediate–they were able to reuse already existing workflows, bringing them available to the new AI Agents. The results? A significant improvement in ticket resolution time.

Learn more about Sema4.ai Actions

Have questions? Get in touch

For technical questions or to start building, check out our documentation