ProductsServicesAI FilmsInsightsAboutContact
{en|de}

AI software glossary: the terms that matter when you build a real product

Plain-language definitions of the AI terms that decide whether a product ships and holds up: RAG, copilots, agents, fine-tuning, embeddings, vector databases, MCP, evals, guardrails, prompt injection, hallucination, and AI-native apps. Written by an in-house engineering studio, not a marketing team.

RAG (Retrieval-Augmented Generation)

RAG is a technique that fetches relevant facts from your own data at query time and feeds them into a language model before it answers, so responses are grounded in your content rather than the model's memory. It is the standard way to make an LLM answer accurately from private or up-to-date information.

Read →

AI copilot

An AI copilot is an assistant embedded inside an existing product that suggests, drafts, or completes work while the user stays in control and approves each step. Unlike an autonomous agent, a copilot keeps a human in the loop — it speeds the person up rather than replacing them.

Read →

AI agent

An AI agent is a system that, given a goal, plans a sequence of steps and uses tools — searching, calling APIs, writing files — to accomplish it with little or no human intervention. Unlike a single prompt-and-answer, an agent loops: it acts, observes the result, and decides the next step.

Read →

Fine-tuning

Fine-tuning is the process of further training an existing language model on your own examples so it adopts a specific style, format, or skill. It changes the model's weights, baking the behavior in, rather than supplying knowledge at query time the way RAG does.

Read →

Prompt injection

Prompt injection is an attack where malicious instructions hidden in content the model reads — a web page, a document, an email, a tool result — trick it into ignoring its real instructions and doing something the attacker wants. It is the LLM equivalent of a confused-deputy attack and the top security risk for AI products.

Read →

Vector database

A vector database stores data as embeddings — lists of numbers that capture meaning — and finds the items most similar to a query by mathematical distance rather than exact keyword match. It is the retrieval engine behind most RAG systems and semantic search.

Read →

Embeddings

Embeddings are numerical representations of text, images, or other data — lists of numbers positioned so that similar meanings sit close together. They let software compare things by meaning instead of exact words, and they are what a vector database stores and searches.

Read →

LLM evals

LLM evals are the tests that measure whether an AI feature actually works — checking the model's outputs against expected results, quality criteria, or human judgement. They are to AI products what unit tests are to ordinary software: the way you know a change helped instead of guessing.

Read →

MCP (Model Context Protocol)

MCP is an open standard for connecting AI models to tools and data sources through a common interface, so the same connector works across different AI apps. Think of it as a universal adapter: build a server once and any MCP-compatible assistant can use it, instead of writing a custom integration per app.

Read →

AI-native app

An AI-native app is software designed around AI from the ground up, where the intelligence is the core of how it works rather than a feature bolted on. The product, data model, and interface assume natural language and model reasoning as primary inputs, instead of adding a chatbot to a conventional app.

Read →

Hallucination

A hallucination is when a language model produces confident, fluent text that is factually wrong or made up — a citation that doesn't exist, a fabricated figure, an invented API. It happens because the model predicts plausible text, not verified truth, and it is the central reliability problem in AI products.

Read →

Guardrails

Guardrails are the rules and checks placed around an AI model that constrain what it can receive and what it can do — filtering inputs, validating outputs, and blocking unsafe actions. They are how you keep an unpredictable model inside safe, on-brand, legal boundaries in a real product.

Read →

Building something that touches these ideas?

Start a project inquiry