Introduction
Luis Marek learned the value of a good prompt while writing applications he did not want to send. The first AI drafts sounded confident, bland and nothing like him. When he changed the question, the answer changed enough to make him curious.
Luis lives in Berlin and supports nonprofit organizations with safe AI workflows. His route includes a dropped business degree, shared-flat debt and many hours in online forums.
Story of the Path into AI
The title “prompt engineer” made some people laugh, and Luis understood why. It can sound like clever typing dressed up as a profession. He had to prove that his work meant more: setting boundaries, testing outputs, forcing source checks and building repeatable instructions.
He learned evaluation methods, privacy basics and knowledge-management workflows. His first project was a set of writing assistants for a counselling centre. The tools helped make letters clearer but were not allowed to provide legal advice. The early prompts were too open. They produced warm, useful-sounding text that quietly invented missing facts.
Luis rewrote them with hard stops: ask for missing information, mark uncertainty, never fill gaps with imagination.
Current Work
Today Luis helps NGOs with communication, grant reports and internal knowledge searches. In one funding application he let AI suggest structure only. The numbers, impact stories and final claims came from the team. That distinction is central to his work.
Small organizations save time without losing their voice to generic text. Luis sees the main risk in convenience. If a tool makes mediocre language easy enough, teams stop noticing when the text no longer carries their own experience.
Personal Advice
“A good prompt is not a trick. It is a clear work instruction with boundaries,” Luis says. He advises beginners to write down not only what the tool should do, but what it must never invent, decide or soften.
Key Facts
Age and place: 23, Berlin.
Background: study dropout, financial insecurity, self-organized learning.
Entry into AI: writing assistants for a counselling centre with strict limits.
Focus today: AI for nonprofit organizations.
Typical tools: prompt design, evaluation, knowledge management.
Werkstattnotiz
Luis keeps failed prompts in a separate document. The worst ones are not absurd; they are charming and overhelpful. He is testing a review step in which teams read AI drafts aloud. Generic language becomes more obvious when it has to pass through a human voice.