Term
Generative AI
Artificial intelligence that creates new content from a prompt: text, images, code, audio or video. Companies use it for writing, summarising, translating or processing documents, but its output must be checked and care taken over what data is fed into it.
Updated
What it means
Generative artificial intelligence is a kind of AI that produces not just predictions or classifications but new content: text, images, audio, video or code. The best known are text models, large language models, which from an instruction, the prompt, write an answer, an email, a summary or a translation.
The models learn from vast amounts of data and produce output that is statistically likely. That is why it sounds convincing even when wrong: a model can invent a law, a section or a quote that does not exist. This is called hallucination and is the main reason outputs must be checked.
Generative AI is quickly moving from a chat tool into everyday software: office suites, email, accounting programs and chatbots on websites. For more complex tasks it is combined with tools to form an AI agent that carries out the task itself.
How it works in practice
We recommend that companies use generative AI for tasks where a person can easily check the output: draft texts, summaries of long documents, translations, preparing tables or answers to routine questions. On tax or legal questions it is a good aid for getting oriented, but an expert must verify the conclusion against current law.
The key issue is data. If employees paste contracts, payroll or clients’ personal data into public AI tools, the company may breach the GDPR and confidentiality. The answer is business licences with contractual safeguards, clear internal rules and, where appropriate, DLP that monitors the input of sensitive data.
Buying AI tools is an ordinary operating expense; developing your own solution may create an intangible asset. Some uses, such as assessing job applicants, are subject to stricter rules under the AI Act.
What to watch out for
The biggest risk is blind trust in the output. A professional-sounding text can contain wrong amounts, non-existent deadlines or outdated rules. Anything going to a client, an authority or into a contract must be checked by someone who understands the subject.
The second risk concerns copyright and confidentiality. AI-generated content may not be protected by copyright and may resemble other people’s work. Contracts with clients and suppliers should address whether and how AI is used in the work and who is responsible for the result.
Need the term explained on your own situation? Consulting from €55 / hour excl. VAT, depending on topic and complexity.