Glossary

AI Terms for Mid-Sized Businesses

Clear definitions of the terms we use most often in our projects — no buzzword bingo, no consultant speak.

Explainable AI (XAI)

AI systems whose decisions are traceable for humans. Instead of a black box, XAI delivers structured explanations of which factors entered the result and how. A core requirement for EU AI Act compliance and for team acceptance.

EU AI Act

The first comprehensive AI regulation by the European Union. In force in stages from 2025, with strict obligations for high-risk AI from 2026. Defines risk classes, transparency requirements, and documentation and audit duties.

• Source: EU Regulation 2024/1689 (official text)

Mittelstand (German SME)

In Germany, companies with 10 to 500 employees, often family-owned. Account for over 99 percent of all companies and more than half of all social-security-contributing employees. Have specific requirements for AI: pragmatic, data-frugal, without large-project character.

• Source: IfM Bonn — Mittelstand definition

Process Automation

Step-by-step replacement of manual work by software or AI. In clean order: first digitize, then automate, then add AI. Without clean data and defined processes, AI runs into the wall of reality.

AI Potential Check

Structured initial analysis, typically a half-day or full-day engagement. Clarifies: where does AI pay off, which use cases have the fastest ROI, which data and processes need to be cleaned up first. Delivers a prioritized roadmap, not a strategy paper.

AI Roadmap

Multi-stage implementation plan: quick wins, pilot projects, scale-up. Includes IT landscape, data state, EU AI Act requirements, and funding options. Goal: not everything at once, but real value step by step.

GDPR-Compliant AI

AI applications that comply with the EU General Data Protection Regulation: lawful processing, right to access and erasure, data minimization, protection of special categories like health or politics. In practice often local or on-premise instead of US cloud.

• Source: GDPR full text (gdpr-info.eu)
• Source: BSI — IT Baseline Protection

Data Strategy

Clarification of which data the company holds, who is allowed to use it, how it is kept clean, and where it lives. Prerequisite for any meaningful analysis, any reporting, and any AI application.

Quick Wins

Small, fast automations with measurable effect in weeks rather than months. Examples from our projects: accounting from 8 hours to 1 hour, quote generation from 3 hours to 30 minutes. Goal: fast successes that build trust for larger AI projects.

Prompt Engineering

Structured techniques for steering AI language models like Claude, GPT, or Gemini. Clear task definition, context, prescribed format, examples. Turns random output into reproducible work results.

On-Premise vs. Cloud

On-premise means AI models and data run in your own data center. Cloud means external providers like AWS, Azure, or OpenAI host the infrastructure. Mid-sized businesses choose based on data sensitivity, cost, and IT capacity.

AI Training (for teams)

Structured staff training: AI fundamentals, GDPR-compliant tool usage, prompt engineering, application within their own area of work. Goal: each function uses AI confidently in daily work — not just operates it.