Getting Started with Your CompanyGPT Demo Account

Welcome to your CompanyGPT demo account! This page walks you through the key concepts in a few minutes and shows you, using pre-built example agents, what you can try right away. For everything beyond that, the full documentation is always available.

Your demo account shows the new chat interface

Your demo account already shows the new chat interface. It is currently only available in the demo access and in pilot environments. Some controls therefore look different from those on some pages of this documentation – for example, the Enter key inserts a line break by default, and you send with Ctrl+Enter (macOS: ⌘+Enter). The New Chat Interface page describes what changes.

What is CompanyGPT?

CompanyGPT is your secure AI assistant for maximum productivity — built around the familiar ChatGPT-style interface, extended with agents, RAG (Retrieval Augmented Generation), and further addons. For a full introduction, see What is CompanyGPT?.

Core Concepts

Chat and Models

You chat as usual and can choose between different AI models depending on your use case and the speed or quality you need. To learn which models are available and what they are suited for, see Model Selection.

Prompts (Prompt Templates)

Prompt templates are saved, reusable requests — independent of model or agent. They’re especially useful for recurring tasks. Learn more under Prompts.

Agents

Agents are advanced AI systems that don’t just respond, but can also use tools to retrieve information — for example, searching your documents or the web. Learn more under Agents.

Knowledge & RAG in Your Demo Account

companyRAG makes large volumes of documents semantically searchable and available to agents — the underlying method is called RAG (Retrieval Augmented Generation).

Note

In your demo account, you cannot create your own collections — we take care of that for you during setup. You can, however, create your own agents that access the existing collections, or use one of the pre-built example agents below right away.

The Pre-Built Example Agents

For your demo account, we’ve prepared four example agents. Just click a link below to start a new chat with that agent directly. Alternatively, you can find the agents in the Agent Marketplace in the sidebar, where they are called [DEMO] Wiki, [DEMO] AI Search Datensatz, [DEMO] AI Search (AI Topics) and Web Research Agent.

Tip

The example data behind the RAG agents (Internal Wiki, Dataset Query, AI Topics) is for illustration only. In a real setup, these agents would access your own documents, datasets, and internal content. The Web Research agent, on the other hand, always works with current, real content from the internet.

Internal Wiki

Answers questions about internal policies, processes, and knowledge — like a colleague who knows your company wiki by heart.

→ Start a chat with the Internal Wiki agent

Sample prompts:

• “How do I submit a vacation request?”
• “What does our home office agreement regulate?”
• “Who do I contact for IT issues?”
• “What does our data protection policy state?”
• “How does the onboarding process for new employees work?”

Dataset Query

Answers targeted questions about a structured dataset, rather than just summarizing individual documents.

→ Start a chat with the Dataset Query agent

Sample prompts:

• “What are the top 10 products by revenue?”
• “How many customers do we have per store?”
• “Show me the average discount per product category.”
• “Compare shipping speed across stores.”
• “Which products currently have low stock levels?”

Generic RAG Agent (AI Topics)

Answers general questions about AI topics based on an indexed knowledge collection — a great way to experience the RAG principle itself.

→ Start a chat with the AI Topics agent

Sample prompts:

• “What’s the difference between RAG and fine-tuning?”
• “Explain what an embedding is.”
• “What is prompt engineering and why does it matter?”
• “What does hallucination mean in the context of AI models?”

Web Research

Searches the internet for current information and summarizes the results with source citations — ideal for researching current topics, companies, or news. Learn more about the underlying principle under Web Search & WebFetch.

→ Start a chat with the Web Research agent

Sample prompts:

• “What are the latest developments regarding the EU AI Act?”
• “Research the most recent press releases from a company of your choice.”
• “What’s new on the topic of GDPR-compliant AI?”
• “Compare current pricing across cloud providers for AI models.”

Creating Your Own Agents

Once you’re comfortable with the basics, you can create your own agents that access the existing collections. See Agents for a guide, and Prompt Engineering for tips on writing good instructions.

Questions?

Your demo account also includes a help agent (CompanyGPT Hilfe) that knows this documentation and answers your questions about CompanyGPT directly in the chat — the fastest way to get an answer.

→ Start a chat with the Help agent

If you want to dive deeper, the full documentation is available starting at What is CompanyGPT?.

Your Next Steps

• Ask the Internal Wiki agent your first question
• Try out the Dataset Query agent
• Test the generic RAG agent on AI topics
• Have the Web Research agent look into a current topic for you
• Create your own prompt template
• Create your first own agent
• Have questions? Ask the Help agent