Automate business processes with AI agents - operated sovereignly? WZ-IT builds and operates automation and AI on your own infrastructure - n8n, local models and gateways, GDPR-compliant from one team. See managed AI servers
Automation is not new - but language models lift it to a new level. Instead of just following fixed rules, AI agents can execute tasks depending on context, use tools and delegate subtasks. This article explains what makes an AI agent, how n8n orchestrates such flows and why sovereign operation matters. As of October 2026.
Table of contents
- From automation to AI agents
- What makes an AI agent
- n8n as an automation platform
- Typical use areas
- Operated sovereignly
From automation to AI agents
Classic automation follows fixed rules: when an email arrives, create a record; when an invoice comes in, forward it. That is reliable but rigid - every new case needs a new rule, and unstructured inputs overwhelm the system.
AI agents add a language model to this automation that derives decisions from the context. The agent does not just follow a rigid flow but chooses steps and tools situationally. This makes automation flexible enough for tasks that previously required human judgment.
What makes an AI agent
An AI agent is more than a model that answers. Three capabilities characterize it:
- Tool use - the agent calls services, APIs or databases instead of just producing text.
- Multi-step work - it breaks a task into steps and works through them, with intermediate results.
- Delegation - complex flows are distributed across specialized agents, such as a research and a service agent.
The difference from simple automation lies in this decision-making ability: not every step is fixed in advance, the model chooses it depending on context.
n8n as an automation platform
For agents to be embedded in real processes, they need a platform that connects them with the company's systems. n8n is a widespread choice for this: a visual low-code platform where you connect apps and services via "nodes" into workflows - largely without code. The source code is available, but n8n is licensed under the Sustainable Use License, a fair-code licence and not an OSI-approved open-source licence: use for your own internal business purposes is free, while enterprise features in .ee files require a separate licence.
Beyond its built-in integrations, n8n also connects tools via the Model Context Protocol: the AI Agent node uses external MCP servers through the MCP Client Tool, and the MCP Server Trigger exposes your own workflows as tools for other agents.
Two properties make n8n interesting for privacy-conscious companies: it originates in Germany and can be fully self-hosted. Combined with a local language model, both the orchestration and the AI decisions run on your own infrastructure. The practical implementation is shown in the guide on business processes with n8n and AI agents.
Typical use areas
AI agents are strong where tasks combine several steps, decisions and access to different systems:
- Sales and marketing - pre-qualify leads, prepare quotes, draft content.
- Customer service - pre-sort requests, answer standard cases, escalate complex ones.
- Data preparation - merge and structure information from different sources.
- Internal processes - bundle research, prepare documents, take over recurring flows.
The benefit does not come from a single magical agent but from combining automation and AI judgment in the right places.
Operated sovereignly
Because agents access company data and internal systems, the operating location is decisive. An agent that passes customer data to a cloud AI shifts exactly the control you want to keep. The sovereign path keeps both in-house: the automation platform self-hosted, the language model local.
Via a gateway like LiteLLM the agents address the models uniformly - sensitive tasks locally, uncritical ones externally if needed. Operation, hardening and maintenance of this interplay can be outsourced as a managed service without giving up data authority.
