Blog
AI is not primarily about technology, but about decisions, says Adastra Czech CEO Štěpán Kopřiva
December 11, 2025
Most companies today are considering how to use artificial intelligence. According to Štěpán Kopřiva, CEO of Adastra Czech, however, the core challenge is not the choice of model, platform, or technology.
In practice, AI initiatives tend to run into decision-making and organizational limits rather than technical ones. Typical obstacles include how companies deal with uncertainty, the state of their data, their ability to connect AI to real business processes, and the expectations they place on AI from the outset.
In the Restart Systému podcast, Štěpán Kopřiva shares experience from early AI projects as well as work with startups and large international enterprises — and explains where AI initiatives most often succeed or fail in reality.
What proves critical in practice
Several recurring principles emerge from the conversation — patterns that appear consistently across industries and types of organizations:
- Without concrete data, AI cannot be evaluated seriously.
Without working with historical and real operational data, it is impossible to reliably assess impact or return on investment. - Uncertainty is an inherent part of AI — and companies must accept it.
Unlike traditional software development, AI does not allow all outcomes to be known upfront. Decisions often have to be made before certainty exists. - Use cases are rarely the problem.
AI typically has plenty of potential applications. The real challenge lies in integrating it into existing environments, connecting it to data, and translating it into a credible business case. - Technology alone does not create value.
What matters is a clearly defined problem and the ability to connect it with data, infrastructure, and the actual way the company operates.
When these conditions are missing, organizations tend to end up with isolated experiments rather than sustainable, long-term change.
Watch the interview (in Czech):
What CEOs and boards actually need to address with AI today
The discussion repeatedly returns to the idea that AI success is not a technical issue, but a management decision. In particular:
- When it makes sense to pursue AI — and when it doesn’t yet.
Without data and a prepared infrastructure, AI cannot be evaluated realistically. - Who owns AI within the organization.
Unclear responsibility between business and IT is a frequent source of friction and failure. - How to manage data quality over time.
Data evolves — and so does AI behavior. This is not a one-off project, but a continuous discipline. - Why small, verifiable steps often work better than large transformations.
Practical experience shows that incremental experiments based on real data carry the lowest risk.
Why the combination of data, AI, and engineering matters
Štěpán Kopřiva also discusses the broader context behind the merger of Blindspot Solutions and Adastra. The key is not AI as a standalone technology, but the ability to combine AI with strong data foundations, engineering capabilities, and real business context.
According to Kopřiva, this combination determines whether AI remains a laboratory experiment or becomes part of how organizations actually operate — across industries such as manufacturing, logistics, telecom, and financial services.
What else you’ll learn in the podcast
- Why companies often demand a precise AI business case before they have the data — and why this approach fails.
- What the reality of enterprise data looked like during early AI projects — in the Czech Republic and abroad.
- Why the main limitation is rarely a lack of ideas, but the ability to bring them into practice.
- How technical excellence differs from the ability to successfully commercialize solutions.
- How the role of the CEO is changing in the age of AI — and why direct contact with customers remains essential.
- What Carlyle’s investment means for Adastra’s future development.
