AI-Powered Workflows

Sovereign Workflows integrates AI capabilities directly into your automation pipeline. Add AI steps to your workflows to summarize text, classify data, extract structured information, generate content, and make decisions — all without leaving the visual editor.

What AI Steps Can Do

AI steps use large language models (LLMs) to process and generate text. Common use cases include:

• Summarization — condense long documents, emails, or reports into key points
• Classification — categorize incoming data (support tickets, leads, feedback) into predefined categories
• Extraction — pull structured fields from unstructured text (names, dates, amounts from invoices or emails)
• Content generation — draft responses, create reports, or generate documentation
• Translation — convert text between languages
• Decision support — analyze data and recommend an action

An AI step takes a prompt (your instructions) and the data from previous workflow steps, sends it to the AI model, and returns the result for downstream steps to use.

Supported AI Providers

Sovereign supports multiple AI providers. You configure which providers are available in your deployment, and each AI step can specify which provider to use.

Configuring an AI Provider

Set the API key for your chosen provider in your environment:

OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-ant-...
AZURE_OPENAI_API_KEY=...
GEMINI_API_KEY=...

You can configure multiple providers simultaneously and choose per-step which one to use. The LLM configuration file controls model selection, parameters, and provider routing.

Using AI in a Workflow

To add an AI step to your workflow:

1. Open the Studio editor
2. From the Node Palette, find the LLM connector
3. Drag a Chat action onto the canvas
4. In the Properties Panel, configure:
   • System prompt — instructions that define the AI's behavior (e.g., "You are a customer support classifier. Categorize the following ticket.")
   • User message — the actual content to process, using expressions to include data from previous steps (e.g., {{ context.data.get_ticket.result.description }})
   • Model — which AI model to use (optional, uses the configured default if not specified)

The AI step returns the model's response as output, available to subsequent steps via expressions.

Structured Output

For many use cases, you need the AI to return structured data — not freeform text. You can instruct the model to return JSON with specific fields:

1. In your system prompt, describe the exact format you want
2. Tell the model to respond only with valid JSON
3. Use a JSON Parse built-in action after the AI step to convert the response into a structured object

This lets downstream steps reliably access specific fields (e.g., {{ context.data.classify.result.category }}) rather than trying to parse freeform text.

AI Steps vs Agent Nodes

AI steps and Agent nodes serve different purposes:

• AI steps run a single prompt and return a result. They are predictable and fast. Use them when you know exactly what you want the AI to do.
• Agent nodes enter a reasoning loop where the AI decides which tools to use and takes multiple actions autonomously. Use them when the task requires judgment and multiple steps. See Autonomous Agent Nodes for details.

License Requirements

AI-powered workflow features require an Enterprise tier license or above. This includes the LLM connector actions and the AI provider integrations. The API keys are yours — Sovereign connects to the providers using your credentials.

Next Steps

• Autonomous Agent Nodes — give AI agents tools to work with
• AI Agent Access (MCP) — let external AI agents invoke your workflows
• Available Connectors — see all available actions including LLM