Yura Deutsch

AI terms, clearly explained

Yura's BotWiki: everything about AI bots and chatbots, from what a chatbot is to automation in daily work. Tips and proven practices.

AI terms, clearly explained: from AI agent to data processing agreement, the terms that turn up in quotes and tenders, in plain language and without the marketing gloss.

AI assistant

Software that carries out clearly defined tasks on request.

Definition

An AI assistant is a software system that carries out a clearly defined task on request: summarising an email, proposing an appointment, answering a question. It works reactively: it waits for input and returns a result.

How it differs from an agent

The difference from an AI agent is the degree of autonomy. An assistant proposes and waits for approval; an agent pursues a goal across several steps and decides along the way. In practice the two blend: many products assist by default and act agentically for selected processes.

On This Page

  • Definition
  • How it differs from an agent

Fundamentals

AI agent

A system that pursues a goal across several steps on its own.

An AI agent pursues a given goal across several steps, selects its own tools and reacts to intermediate results. Example: “find a slot everyone can …

Large language model (LLM)

A language model trained on very large amounts of text.

A large language model is trained on very large amounts of text and computes the most likely continuation of an input. From that come its …

Prompt

The input that instructs a language model.

The prompt is the input that instructs a language model. It carries the task, often additional context (documents, conversation history) and requirements for the form …

Hallucination

An invented but plausible-sounding statement from a model.

A hallucination is a statement a model produces that has no basis in the underlying data. Typical cases: invented case numbers, non-existent legal provisions, wrong …

Knowledge base

The body of information an AI assistant draws on when it answers: the knowledge of your business, not of the internet.

A knowledge base is the body of information an AI assistant draws on when it answers: opening hours, processes, product details, standard replies. It is …

AI in business

Workflow automation

Recurring steps run without manual involvement.

Workflow automation means recurring steps run without manual involvement: a trigger occurs, the system performs defined steps, and the result lands where the work continues.

AI phone assistant

Answers calls automatically, recognises the request and acts.

An AI phone assistant answers incoming calls automatically, turns speech into text, recognises the request and responds: giving information, booking an appointment, noting a callback …

Intent recognition

Determining the intention behind an utterance.

Intent recognition determines the intention behind an utterance. “I need an earlier slot”, “would next week work?” and “can we move it?” all aim at …

RAG (retrieval-augmented generation)

The model answers from documents retrieved beforehand.

With RAG , the model answers a question not from memory but from documents retrieved for that purpose. The sequence: take the question, search the …

Human in the loop

A person approves before an action takes effect.

Human in the loop describes processes in which a person checks and approves at a defined point before an action takes effect outside the company. …

KPI (key performance indicators)

Figures that show whether a goal is being reached, such as response time or the share of calls answered without waiting.

A KPI (key performance indicator) is a figure that shows whether a goal is being reached, for example the response time to inquiries or the …

Cohort

A group that shares a starting characteristic. Cohorts are compared over time to read developments cleanly.

A cohort is a group of people or cases that share a characteristic at the same point in time, for example all customers who started …

Privacy & law

GDPR

The European General Data Protection Regulation.

The General Data Protection Regulation governs across Europe the conditions under which personal data may be processed. For AI applications four points matter most: legal …

Data processing agreement (DPA)

The Art. 28 GDPR contract between controller and processor.

The data processing agreement under Art. 28 GDPR sets out what a service provider may do with its client's data. It defines purpose, duration, categories …

EU AI Act

The EU's AI Regulation; substantive obligations apply since 2 August 2026.

The EU AI Act is the European Union's regulation on artificial intelligence. It classifies systems by risk and attaches different obligations to each class. The …

Data residency & EU hosting

Where data is stored and processed, and which law then applies.

Data residency describes the country in which data is stored and processed. That location helps determine which level of protection applies and which authorities can …

Anonymisation & pseudonymisation

Two protective measures of very different strength.

With pseudonymisation , identifying details are replaced by a token but can be restored with additional knowledge. The data stays personal data. With anonymisation the …

Technology

API

A defined interface through which programs exchange data.

An API (application programming interface) is a defined interface through which two programs exchange data. It describes which requests are allowed, what information they need …

Webhook

A notification a system sends on its own when an event occurs.

A webhook is a message a system sends to a stored address when an event occurs, instead of waiting to be asked. Example: as soon …

Audit log

A continuous record of who did what and when.

An audit log continuously records which action was triggered when, and by whom or by which system. It exists for traceability rather than for debugging …

Book a demo