Practical AI Workshop in Târgu Mureș - May 5, 2026

AI Workshop in Hungarian – When a 4-hour workshop turns into almost 6

On May 5, 2026, we organized the second event in the new series of Dalbe AI workshops, at Metamorfozum Concept, in Târgu Mureș.

It was the second workshop in the new series. The first event took place a week before, during the day. This time we intentionally chose an evening slot, so that those working during the day could also attend.

The workshop started at 17:00.

And it ended almost at 23:00.

Not because we were late.

But simply because the conversations wouldn't stop.

Very quickly, the event was no longer just a workshop where people sit and watch a slideshow.

It became a space where we discussed:

• real webshops,
• marketing workflows,
• automation ideas,
• SEO problems,
• and concrete business questions.

Towards the end of the evening we were already discussing:

• AI workflows,
• content generation,
• webshop structures,
• and real ways these tools can be integrated into work.

The workshop didn't end because we ran out of topics.

It ended because people realized they still had to go to work the next day.

And because someone had to pack up the cables and extension cords at the end. AI still doesn't do that by itself. For now.

Who attended?

The attendees included:

• marketing specialists,
• webshop owners,
• photographers,
• former Dalbe clients,
• and people working daily in the digital field.

We were happy to see Sikódi Alfréd again, for whom we built the Clothingoo webshop about two years ago.

Also attending was Ny. Cs., who manages a medical devices webshop, as well as Módi Zsuzsanna, founder of Framed.ro, specializing in family and baby photography.

Because most attendees were already working with:

• websites,
• online marketing,
• webshops,
• or content creation,

the discussions quickly moved towards practical examples and real workflows.

AI Basics – before the whole internet called everything "artificial intelligence"

The workshop didn't start with prompting.

We started with the question:

• what AI is,
• what an LLM is,
• and what actually happened in the last few years.

Because right now the internet calls almost anything "AI".

In reality, much of the current hype is related to LLMs – Large Language Models.

One of the simplest examples we used was the cat example.

If a system has seen thousands of images of cats, and then we show it a new image, the system doesn't "understand" that it sees a cat.

It calculates the probability that that image resembles the patterns it previously learned, and so the "understanding" becomes a mathematical calculation.

This helped many participants understand:

• why AI makes mistakes,
• why it hallucinates,
• and why it shouldn't be blindly trusted.

DeepMind and the beginning of the LLM revolution

We also discussed the 2017–2018 period, when teams like Google DeepMind started building the foundation of the modern systems we use today.

From here we got to:

• OpenAI,
• Anthropic and Claude,
• Google Gemini,
• DeepSeek,
• and the AI projects associated with Elon Musk.

We didn't discuss these systems from the hype perspective.

But practically:

• where they are useful,
• where they have limits,
• and in what types of workflows they work well.

ChatGPT, Claude, Gemini, and the other systems

One of the most important takeaways from the workshop was that these tools are not identical.

ChatGPT

ChatGPT remains one of the most powerful tools for:

• general tasks,
• structuring,
• brainstorming,
• copywriting,
• and daily workflows.

Here we also discussed:

• privacy settings,
• data management,
• connectors,
• and the importance of responsibility in use.

We even analyzed:

• chat,
• image generation,
• deep research,
• learning mode,
• and agentic workflows.

An interesting part of the workshop was using JSON-based prompting.

For many participants, this was the moment they realized:

• how structured you can work with AI,
• and how much the structure of a prompt matters.

 

Claude and Claude Cowork

Claude stood out in particular for:

• handling long documents,
• planning,
• automation,
• and complex thought processes.

Claude Cowork was one of the biggest surprises of the evening.

When we presented a real workflow example for software ordering and process organization, the discussion quickly turned to:

• project management,
• automation,
• and AI used as a working partner.

For many attendees, this was the moment when AI no longer felt just like a chatbot.

But a digital collaborator.

Google Gemini, NotebookLM and Google AI Studio

With Google Gemini, we specifically discussed integration with the Google ecosystem.

Then we moved to one of the most appreciated parts of the workshop:

• NotebookLM.

The moment we started using:

• documents,
  
• PDFs,
• and real source materials,

the atmosphere in the room visibly changed.

Curiosity quickly turned into:

• "ok, this is actually useful".

We explored:

• audio summaries,
• video outputs,
• infographics,
• Google AI Studio,
• and VEO 3.1.

all via NotebookLM.

Prompting – the part most people underestimate

One of the most important sections of the workshop was prompting.

Because most people still use AI like Google.

In reality, the difference between a poor result and a very good one often comes from:

• context,
• roles,
• and prompt structure.

We discussed:

• tokens,
• context windows,
• prompting frameworks,
• no-shot, one-shot and multi-shot prompting,
• COSTAR and SMART frameworks.

One of the important conclusions for the participants was:

• the main limit is not the AI,
• but the way we instruct it.

Generating images, music and content with AI

Creative workflows were not missing either.

We explored:

• Nano Banana 2,
• image generation workflows,
• JSON templates for images,
• and SUNO.

In the SUNO example:

• we generated lyrics with ChatGPT,
• then we created music using SUNO.

Yes. Part of modern business workshops in 2026 means people generating music with AI at 21:30 while discussing workflows and automations.

Interesting times.

We also discussed:

• Gamma AI,
• Lovable,
• design generation,
• and creative workflows based on brand guides.

SEO and AI – a mini SEO workshop for the online area

Because most participants worked in the online field, we obviously got to SEO as well.

We discussed:

• meta titles,
• meta descriptions,
• heading structures,
• ALT tags,
• and their importance.

We also showed:

• how AI can help with SEO structuring,
• generating keyword ideas,
• and organizing webshops.

But we also insisted on an important point:

AI does not replace expertise.

It amplifies it.

If someone builds the wrong SEO strategy, AI will help them produce wrong SEO faster and on a larger scale.

Technology scales human mistakes very efficiently too.

The end of the workshop – brainstorming, voice AI and real questions

The last part of the evening turned almost completely into:

• Q&A,
• brainstorming,
• voice AI workflows,
• and discussions about real business problems.

And honestly?

This was one of the most valuable parts of the whole evening.

It wasn't just a presentation.

It was shared thinking.

It's the kind of thing that's very hard to replicate in a simple online video.

Because when:

• a webshop owner,
• a marketer,
• a photographer,
• and people working in digital projects

start discussing the same workflow, practical ideas and real connections emerge very quickly.

Why do we organize such workshops?

Because the real difference in the coming years will not be:

• who uses AI.

But:

• who uses it well.

And many people:

• are still afraid of it,
• use it wrong,
• or simply don't know where to start.

The goal of Dalbe workshops is exactly this:

• practical examples,
• clear explanations,
• real workflows,
• and tools that can actually be used at work the next day.

Because AI is no longer science fiction.

It has become part of the modern work environment.