Interview: Petr Kameník – AI is our reinforced concrete

How do you think AI has changed in recent years – and where is it all heading?

AI is no longer just the domain of tech enthusiasts. Today, it is becoming a common part of the workday – even though many people are not fully aware of it yet. The number of tools is increasing at a dizzying pace. Everyone promises something, but very few actually deliver.

I have been following this field for a long time and during that time I have developed one analogy – AI is like reinforced concrete to me.

Reinforced concrete and AI? That sounds a bit strange...

It may sound strange, but it is a metaphor based on a true story. In the 19th century, a French gardener wanted to build a durable planter. He tried mixing cement with iron rods – and it held. By chance, he discovered a material that became the foundation of modern architecture a few decades later.

Today, it is similar with AI. We are still experimenting with individual things – generating something here, testing something there. But I feel a potential in it that could fundamentally change the way we work across industries. We just don’t know exactly when that breakthrough will come.

Looking back – when did we at PUX first realize that AI is not just an add-on, but something that can transform the way we work?

At first, we viewed AI as an assistant. Something that would make our work easier – like speeding up translations or helping with content summarization. But the more we worked with it, the more we began to realize that AI is not just a faster Google with text output.

At one point, we said that it has potential for more complex processes. And that we could try to create something of our own. A tool that would not depend on a specific UI or service, but that we could assemble according to our needs. That’s how Balzaq was created.

And the first impulse? Did it come from the client, or was it more our own search for a solution?

It was a quite specific request. The client TASY needed to generate hundreds of product texts and ideally right away – additionally in multiple languages. Classic manual work would not suffice. And that was the moment we said: let's fully utilize AI. We started automating.

Previously, copywriters could write about thirty articles a day. And that was only in one language. Today we can handle 300 articles daily. Easily in five variations.

What exactly can Balzaq do? How would you describe it to someone who has never heard of it?

It is an automation tool.It is not a one-purpose generator, but a system designed for specific repetitive tasks that companies deal with every day – often handling them manually. For example, product descriptions, translations into several languages, categorization, sorting, tagging, or matching with data.

Balzaq can work with data automatically, but also in a structured and intelligent way. It reads information from Excel files, XML, or APIs, provides context, generates content according to predefined rules and formats – while allowing for fine-tuning. You can adjust the outputs according to brand style, communication tone, or target audience.

We assembled all of this from things we knew well and were already using. We just put it together in a way that saves time and can be easily integrated into real processes.

How have we progressed since then?

We have continued to advance Balzaq technically – we started with text generation, but soon realized it could do much more. We are doing things I would have previously considered science fiction. For example, OCR – recognizing text from images. In one project, we needed to read information from a tire in a photo.

However, those photos are not ideal – the tire is dirty, angled, in shadow… and regular AI struggles with that. It’s not enough to just throw it into one model and expect perfect results.

So we have three AI components working in sequence. One finds the wheel, the second "flattens" it, and the third reads what it has. Finally, it checks against a database – to see if those details should actually be there. In other words: it verifies that what the AI read is not nonsense, but a genuinely existing value. This is an important step, as this is often where the reliability of the entire system is revealed.

So no universal AI magic?

Exactly. Every project has different requirements, different data, different goals. It's not about having one tool that solves everything.

At Balzaqu, this is even more true. We always choose which components and models to use – and we calculate everything carefully. Sometimes you don't need the absolute best possible output. You need an output that is worth it. And that is a completely different way of thinking than just "what technology can do."

How has the arrival of artificial intelligence changed the work of people at PUX?

A lot. A copywriter today is not just someone who writes. They must understand prompting and be able to communicate with AI to get quality output.

A graphic designer must be able to prepare multiple variants, respond quickly, and think about what makes sense. And actually, every team member must better understand what they do and why. Because AI can easily produce something that looks great at first glance, but it could be complete nonsense.

Importantly, though, AI does not replace anyone here. It only shifts the way we think about work. It doesn't mean you don't need a copywriter – it means the copywriter must know other things than before. And this applies to all professions in the team.

So even though we actively use AI today, is caution still warranted?

Definitely. A tool is not a solution. It's not that when we have AI, everything will run on its own and without errors. Critical thinking, verification, and testing are still necessary. And above all, understanding what the tool actually does and why.

It's like with reinforced concrete. You can build a bridge with it – strong, functional, meaningful. Or you can make a crooked flower pot that will crack soon. Technology alone is not enough – it depends on how you use it.

What are the future plans for Balzaq?

We might open it up more. Custom UI is nice, but not for everyone. Rather, we are thinking about how to connect it to the tools that people are already using. And provide them with our scenario.

What would you advise companies that are just observing AI for now?

Try it out. But do so wisely. Don't dive in just because others are doing it. Determine what you actually want to address. And expect that the first phase will be excitement. The second will be disappointment. And only the third will work.

AI is not magic. It is a technology that is here to stay. Today it may seem like a few boxes. But in a few years? It will be hard to build without it.

Are you interested in how AI could work in your project? Or are you wondering if you should even get started?
Contact us – we would be happy to discuss where it makes sense to begin.