Skip to main content

Glossary

Your guide to AI, technical, and Swiss regulatory terminology used throughout this documentation.

AI & Machine Learning Terms

LLM (Large Language Model)

An artificial intelligence system trained on vast amounts of text data to understand and generate human-like text. Examples include the models available on Schatzi AI (Thinking, Standard, etc.). LLMs can perform tasks like writing, analysis, translation, and summarization.

Tokens

The basic units that AI models use to process text. Roughly:

  • 1 token ≈ 4 characters in English
  • 1 word ≈ 1-2 tokens on average
  • 1 page of text ≈ 500-600 tokens

Tokens are how AI usage is measured and billed. Both your input (prompt) and the AI's output (response) consume tokens.

Context Window

The maximum amount of text (measured in tokens) that an AI model can consider at one time. This includes your prompt, any uploaded files, and the AI's response. Different models have different context windows:

  • Standard: 32,000 tokens (~24,000 words)
  • Thinking Mini: 64,000 tokens (~48,000 words)
  • Document Analysis: 128,000 tokens (~96,000 words)
  • Thinking – Large: 200,000 tokens (~150,000 words)

Prompt

The instruction or question you give to an AI model. A well-crafted prompt provides context, specifies the task, and defines the desired output format. Quality of prompts directly affects quality of responses.

Completion

The AI model's response to your prompt. Also called "output" or "generation."

System Message

Instructions that set the overall behavior and role for the AI model (e.g., "You are a Swiss financial analyst"). These guide how the model approaches all tasks in a conversation.

Parameters

Settings that control how an AI model generates text:

  • Temperature: Controls randomness (higher = more creative/varied, lower = more focused/deterministic)
  • Max Tokens: Limits the length of the AI's response
  • Top-p: Alternative way to control randomness

Latency

The time delay between sending a prompt and receiving the AI's response. Faster models (like Standard) have lower latency. More powerful models (like Thinking – Large) may have higher latency as they perform more complex reasoning.

Fine-tuning

The process of further training an AI model on specific data to specialize it for particular tasks or domains. Note: Schatzi AI does not fine-tune models on customer data, ensuring zero data retention.

Training Data

The text and information used to teach an AI model. AI models learn patterns from this data. Important: Your prompts and files are never added to training data on Schatzi AI.

Hallucination

When an AI model generates information that sounds plausible but is actually incorrect or fabricated. Always verify important facts, especially names, dates, statistics, and technical details.

Embedding

A numerical representation of text that captures its meaning, allowing AI systems to understand semantic similarity. Used in search and retrieval systems.


Data Privacy & Compliance Terms

FADP (Federal Act on Data Protection)

Switzerland's primary data protection law, revised and strengthened in 2023. FADP governs how personal data must be handled, stored, and protected in Switzerland. Key principles:

  • Lawfulness and good faith in processing
  • Purpose limitation
  • Data minimization
  • Accuracy
  • Storage limitation
  • Security and confidentiality

GDPR (General Data Protection Regulation)

The European Union's comprehensive data protection regulation. While Switzerland has its own laws (FADP), GDPR is relevant for:

  • Swiss companies doing business in the EU
  • Comparison and harmonization with Swiss standards
  • International best practices

Data Processor

An entity that processes personal data on behalf of a data controller. In the Schatzi AI context, Schatzi acts as a data processor when customers upload personal data.

Data Controller

The entity that determines the purposes and means of processing personal data. When you use Schatzi AI, you typically remain the data controller of any data you process.

Personal Data / Personal Information

Any information relating to an identified or identifiable natural person. Examples:

  • Names and contact information
  • Identification numbers (passport, social security)
  • Location data
  • Online identifiers (IP addresses, cookies)
  • Financial information
  • Health data

Data Sovereignty

The principle that data is subject to the laws and governance of the country where it's physically located. Schatzi AI provides Swiss data sovereignty – your data is stored and processed only in Switzerland, subject to Swiss law.

Zero Data Retention

Schatzi AI's policy of not storing customer prompts, files, or outputs longer than technically necessary. Your data is not used to train models or retained in our systems.

Anonymization

The process of removing or modifying personal identifiers from data so individuals cannot be identified. Best practice when possible before using AI.

Pseudonymization

Replacing identifying information with pseudonyms or codes. Provides more privacy than raw data but is reversible, unlike anonymization.

Data Breach

Unauthorized access, disclosure, alteration, or destruction of personal data. Organizations must report significant breaches to authorities and affected individuals under FADP and GDPR.


Swiss Regulatory Terms

FINMA (Swiss Financial Market Supervisory Authority)

Switzerland's independent financial markets regulator. Supervises banks, insurance companies, stock exchanges, and other financial institutions. Issues guidelines and enforces compliance with Swiss financial regulations.

SNB (Swiss National Bank)

Switzerland's central bank, responsible for monetary policy and financial stability.

Banking Secrecy

Swiss legal principle protecting confidentiality of bank-client relationships. While modified in recent years for tax transparency, Swiss privacy protection remains strong.

KYC (Know Your Customer)

Regulatory requirement for financial institutions to verify client identity and assess risk. Part of anti-money laundering (AML) compliance.

Cross-Border Data Transfer

Moving data from one country to another. Under FADP, data can only be transferred to countries with adequate data protection. Using Schatzi AI keeps data within Switzerland, avoiding cross-border transfer complexities.


Technical Terms

API (Application Programming Interface)

A way for different software systems to communicate. Schatzi AI is accessed through OpenWebUI, which uses APIs to connect to the AI models.

Cloud Computing

Delivery of computing services (storage, processing, software) over the internet. Schatzi AI uses Swiss-based cloud infrastructure.

Encryption

Converting data into coded form to prevent unauthorized access. Data should be encrypted both:

  • In transit: While being sent over networks
  • At rest: While stored on servers

GPU (Graphics Processing Unit)

Specialized computer processor originally designed for graphics but now essential for AI computations. Schatzi AI uses NVIDIA H100 GPUs for high performance.

Inference

The process of using a trained AI model to make predictions or generate responses. When you ask Schatzi AI a question, the model performs inference to generate the answer.

Infrastructure

The underlying technology foundation including servers, networks, and data centers. Schatzi AI's infrastructure is located entirely in Switzerland.

Model

In AI context, the trained system that performs tasks. Schatzi AI offers multiple models optimized for different use cases (Standard, Thinking, Data Analysis, etc.).

On-Premises vs. Cloud

  • On-Premises: Software/systems hosted on the organization's own servers
  • Cloud: Software/systems hosted by a service provider

Schatzi AI is cloud-based but with Swiss data sovereignty.

SaaS (Software as a Service)

Software delivered over the internet on a subscription basis. Schatzi AI is a SaaS platform.

Server

A computer system that provides services, data, or resources to other computers ("clients") over a network.

SSL/TLS (Secure Sockets Layer / Transport Layer Security)

Protocols for encrypting data transmitted over the internet. Ensures secure connections (look for https:// in URLs).


Business & Financial Terms

ROI (Return on Investment)

Measure of profitability: (Gain from Investment - Cost of Investment) / Cost of Investment. Used to evaluate if Schatzi AI subscription delivers value.

KPI (Key Performance Indicator)

Measurable value showing how effectively objectives are being achieved. Examples for AI usage:

  • Time saved per week
  • Tasks completed
  • Cost per task
  • Quality improvements

Use Case

A specific situation or scenario where a technology or solution is applied. This documentation includes use cases for asset management, retail, and professional services.

Workflow

A sequence of steps to complete a task or process. This documentation provides workflows for common business tasks using AI.

SME (Small and Medium-sized Enterprise)

In Switzerland, typically companies with:

  • Small: <50 employees
  • Medium: 50-250 employees

Schatzi AI is designed for Swiss SMEs' needs.

COGS (Cost of Goods Sold)

Direct costs of producing goods or services sold. Used in financial analysis and reporting.

EBITDA (Earnings Before Interest, Taxes, Depreciation, and Amortization)

Measure of company profitability. Common metric in management reporting.


Schatzi AI Specific Terms

Model Profiles

The different AI models available on Schatzi AI, each optimized for specific tasks:

  • Standard: General purpose, fast
  • Thinking Mini: Balanced reasoning and speed
  • Thinking – Large: Complex analysis and reasoning
  • Data Analysis: Spreadsheet and structured data
  • Document Analysis: PDFs and long documents
  • Web Search: Current information access

Plan Tiers

Schatzi AI subscription levels:

  • FREE: Trial access to basic models
  • LIGHT: Essential models for regular users
  • BASIC: All models including Web Search

Swiss Green Energy

Schatzi AI infrastructure runs primarily on renewable/green energy sources, consistent with Switzerland's commitment to environmental sustainability.


Usage Tips

When You See an Unfamiliar Term

  1. Check this glossary first
  2. Search the documentation
  3. Contact support for clarification
  4. Ask the AI model itself to explain (it can define technical terms!)

Keep Learning

AI technology and terminology evolve rapidly. We update this glossary as new terms become relevant to Schatzi AI users.


Ask the AI

If you encounter a term not in this glossary, try asking one of the Schatzi AI models to explain it! For example: "Explain what [term] means in the context of AI and data privacy."