Data scientist Farkas Dávid, co-founder of Principle Zero, argued in a podcast that artificial intelligence is often mistakenly treated as a universal solution for corporate problems. Many companies assume that introducing an AI tool will automatically improve efficiency, speed up work, and lead to better decisions. According to Farkas, this is a dangerous oversimplification.
He frames AI primarily as an amplifier: it strengthens the processes and behaviors already present in an organization. If a company operates consciously, has well-designed digital processes, understands its own data, and knows where it wants to create value, AI can provide a significant competitive advantage. Conversely, if a company merely digitizes its analog processes and then overlays AI, it risks reproducing the same mistakes faster and at scale.
Organizational and human factors are decisive
Farkas says that migrating a company to AI is only partly a technical issue. IT security, infrastructure and technical implementation must be solved, but the larger challenge is human and organizational: companies need to understand how people work, where decision points are, what can be automated, and what should not be entrusted to machines. That is vital because current AI systems do not think for us; they amplify existing decision-making — for better or worse.
Problems highlighted in the conversation include poor data management, imprecise workflows and inadequately thought-through business logic, all of which can be magnified by inappropriate AI use.
Startups and durable business value
For AI startups, the order matters: first find a real problem, then evaluate whether AI is the right solution. Sometimes a simpler automation, a well-designed process, or even a manual tool will be more effective. Farkas warns of the risk when companies build only a thin layer on top of a large language model: if a startup takes a frontier model and adds only a small interface or a couple of extra features, a subsequent model update could erase its business value in one move.
Durable value, he argues, comes not from API calls but from building proprietary knowledge bases, processes, logic and market experience. A solid AI process must perform consistently across many runs: firms must know what information goes in, what format responses arrive in, how results are verified, and how the next step is constructed. That is no longer a one-off demonstration but a real business system.
Not everyone competes for frontier models
On the global AI competition, Farkas points out differences between approaches: the United States pursues the most powerful models, while China and Europe may focus more on cheaper, more open or enterprise-friendly ways to deploy the technology. The example of Mistral shows that not every player wants or can enter the capital-intensive frontier-model race, so emphasis may shift to implementation and operability.
What if every office worker had a premium AI assistant?
The podcast also asked what would happen if every Hungarian office worker suddenly received an unlimited premium AI assistant. Farkas believes many would likely not use it — a new tool by itself does not change how work is done. Motivation, training, managerial intent and clear goals are needed for AI to actually improve work. Without control, AI can create absurd situations and inefficiencies: one person may ask the AI to write long emails from five thoughts, while another uses it to compress the same five thoughts into a short note.
Practical takeaway
AI becomes valuable not by being added everywhere, but by applying it where there is a clear need, understanding where it is risky, and keeping human judgment where it must remain. The podcast episode with recurring guest Aczél Petra, communication researcher and professor at Széchenyi István Egyetem, covers a range of related topics with time markers for listeners.
Selected timestamps from the conversation:
- What connects understanding human thought to data? (01:01)
- Difference between data and information, and how AI could filter information noise (05:16)
- What it means to be a data scientist today and how the role has changed (06:24)
- Can AI replace the work of a data scientist? (09:10)
- Can a company exist without AI today, and what does the label “AI company” really mean? (13:52)
- What gives lasting business value to an AI-based service? (15:31)
- How to found an AI startup in 2026? (20:32)
- Differences between US, Chinese and European AI strategies (22:57)
- When does AI become dangerous for a company? (28:33)
- What if every Hungarian office worker had a premium AI assistant? (32:59)
The episode’s main message is that AI’s business value depends not on the technology alone but on how organizations reshape processes, data use and decision-making so the technology amplifies strengths rather than reproducing flaws.
