Knowledge Agents: The Secure Internal ChatGPT for Your Employees

Every SME has a knowledge problem: information is buried in emails, SharePoint folders, PDFs, and in the heads of individual employees. When someone leaves or goes on vacation, knowledge is lost. An internal knowledge agent solves this problem - securely, privacy-compliant, and available 24/7.

What is a knowledge agent?

A knowledge agent is an AI system trained on your internal documents, processes, and FAQs. Employees can ask questions in natural language - the agent searches your documents and returns a clear answer. Think of it as an internal ChatGPT that only knows your company's knowledge.

The difference from ChatGPT

• Privacy: Your knowledge agent runs on Swiss infrastructure. No company data flows to OpenAI, Google, or Microsoft.
• Accuracy: The agent answers only based on your documents - no hallucinations from internet training.
• Access rights: Different employees see only the information they're authorized for. HR documents remain invisible to non-HR.
• Source references: Every answer includes a link to the original document. Employees can verify the source.

Typical use cases

• Onboarding: New employees ask the agent instead of disturbing colleagues. Questions like 'Where do I find the expense policy?' or 'How do I request vacation?' are answered instantly.
• IT support: 'How do I connect the VPN?', 'My printer isn't working' - the agent knows all IT guides and provides step-by-step help.
• Product knowledge: Sales staff ask the agent for technical specifications, price lists, or competitor comparisons.
• Compliance: 'Can I send this customer data by email?' - the agent knows your internal policies and gives an immediate answer.

The technology behind it: RAG explained simply

RAG stands for Retrieval-Augmented Generation. The principle: before the AI generates an answer, it first searches your document collection for relevant passages. These passages are then provided as context to the language model, which formulates a clear answer.

The advantage: the language model doesn't need to be trained on your data. New documents are indexed within minutes and immediately available for questions.

Swiss Hosted: Why it matters

Public AI tools like ChatGPT or Google Gemini process your inputs on US servers. For sensitive business data, this is problematic - especially under the FADP. A Swiss Hosted knowledge agent uses Swiss data centers and open-source language models that run locally. Your data never leaves Switzerland.

What does it cost?

• Setup and configuration: From CHF 8,000 (incl. document import, access rights, testing)
• Running costs: CHF 200-500/month (hosting + model inference)
• ROI: Typically 2-4 hours per employee per week saved

With 30 employees and an hourly rate of CHF 80, that's CHF 20,000+ per month in productivity gains.