As AI agents become increasingly capable, a key question emerges: how can companies ensure they operate safely, reliably and effectively in a business environment? This topic came up repeatedly in discussions at Data Festival 2026 at Prague University of Economics and Business. We share our takeaways from the event and explore the challenges companies face when adopting AI.
On Tuesday, November 3, we once again attended Data Festival at Prague University of Economics and Business. As expected, the capabilities of artificial intelligence and AI agents were among the main topics. But far more interesting than what these technologies can already do was the question of whether companies are truly ready to deploy them in production.
Several key themes came up repeatedly throughout the discussions.
Semantic layer: Without context, AI agents will waste time and resources
An AI agent may have access to vast amounts of data. But if it doesn't understand what that data means or how it relates to other information, it will resort to trial and error to find answers.
This can lead to dead ends, incorrect interpretations and unnecessarily wasted tokens.
That's why topics that may seem less exciting than generative AI are once again coming to the forefront, even though they are essential to making it work reliably: ontologies, data catalogs, consistent definitions and machine-readable knowledge bases.
Governance and ownership: Every agent needs clear accountability
Who is responsible for ensuring that an agent provides accurate answers? Who decides which data it can access? Who approves changes to its behavior? And who is accountable if it causes harm?
Without clear answers to these questions, it is difficult to consider an agent truly governed and production-ready. Instead, it remains an experiment without clear ownership or defined accountability.
Moving to production: A demo is not a product
We've seen countless impressive demonstrations of AI agents. But the real challenges begin when it's time to deploy an agent into a production environment.
Integrations, access permissions, security, monitoring, auditability, output quality, costs and incident response all need to be addressed. And that's not something one person can solve in a few evenings of vibe coding.
Deploying AI into production is therefore a complex undertaking that requires a systematic approach and collaboration among experienced IT and data teams, many of which are already operating at the limits of their capacity.
Clear boundaries: What should an agent actually be allowed to do?
An agent that simply recommends the next step poses a different type of risk than one that can independently modify production data, place orders or trigger business processes.
As autonomy increases, so does the need for clearly defined rules, control mechanisms and human oversight. The question is not just what an agent can do, but also what a company can safely entrust it to do.
From experiments to real business value
And these are not just theoretical questions raised at Data Festival. We encounter similar challenges with our clients every day. Companies want to harness the potential of AI, but even more importantly, they need to build reliable data foundations, establish governance, ensure security and manage the transition to production.
Fortunately, they don't have to tackle these challenges alone. At UD4D, we help companies turn AI experiments into reliable, practical solutions that deliver value not just in demos, but above all in everyday business operations.
Frequently asked questions
What were UD4D's key takeaways from Data Festival 2026?
One of the main takeaways was that AI agent capabilities are advancing faster than many companies' readiness to deploy them in production. Beyond the capabilities of AI itself, increasing attention is being paid to data context, accountability, security and long-term management.
How can companies assess their readiness to deploy AI agents?
A good starting point is to assess whether the company has reliable, clearly defined data, appropriate access permissions, clearly assigned responsibilities and processes for validating AI outputs. It is also important to be able to monitor agent operations and intervene safely if an agent does not behave as expected.
What role does data governance play in AI adoption?
Data governance helps organizations define what data is available, what it means, who is responsible for it and how it can be used. For AI agents, this provides a foundation for working with trustworthy information within established rules.
