On 15 April 2026, from 10:00 to 11:30, Viktor Csuvik presented the online webinar “OpenAI modellek a gyakorlatban: beállítások, döntések, kockázatok” (OpenAI models in practice: settings, decisions, risks) for Verlag Dashöfer, in Hungarian. It was aimed at executives, IT managers and digital and innovation leads.
What it covered:
- Why AI's business value depends on how it is used, not on how “smart” the model is
- Key settings such as system instructions, token limits and temperature, and why technical settings are really coded business rules
- Temperature as a business risk
- Why long prompts become unstable and expensive
- Lessons from real OpenAI projects, including why the last 20% is the hardest
- Choosing a model by business value
- Turning a demo into a reliable system
