PAMETNA HRVATSKA Sandra Catovic Strucic, powered by Janaf

Guest Sandra Čatović StručićHosted by Boris ŽivkovićMay 3, 2026

Sandra Čatović Stručić

About this episode

What you'll hear in this conversation

Every new technology brings the promise of efficiency, but also the danger of overreach. ChatGPT, for all its capabilities, has explicitly stated it won't give legal or medical advice. Yet, the legal profession is drowning in a "forest" of regulations, from EU directives to national laws, where finding precise answers is critical. This is the tension Sandra Čatović Stručić and her co-founders at Ulpian set out to resolve. A lawyer herself, Sandra co-founded Ulpian with a team of natural language processing experts, building a specialized generative AI tool that cuts through the regulatory burden without sacrificing accuracy. You'll hear why general AI tools fall short in high-stakes fields, and how combining deep domain knowledge with advanced technology can create clarity in complex environments.

Insights from the conversation

What to take from this episode

Ulpian's four co-founders represent an equal split between legal experts and AI specialists who have been working with natural language processing for a decade. Building a specialized AI for a highly regulated field is not about pairing two disciplines; it demands a deep, inseparable fusion of long-standing technical expertise with profound domain knowledge.

ChatGPT recently updated its terms and conditions to explicitly state it will not provide legal or medical advice. This explicit refusal from general AI models signals that for high-stakes, regulated domains, the responsible path forward lies in developing specialized tools, rather than attempting to adapt general ones.

Sandra uses ChatGPT for simple tasks like finding an "easier goulash" recipe, but warns against patients self-diagnosing from Google or lawyers receiving "catastrophic" contracts from general AI. The convenience of general AI for trivial matters can obscure its severe risks when applied to critical, specialized areas, where the tolerance for error is effectively zero.

The EU alone has thousands of regulations, which then multiply across member states, creating a "very large regulatory burden" complicated by directives, national laws, and varied implementation. Navigating this intricate web of legal instruments demands an AI that understands the layered relationships and specific application of each provision, not merely a search engine that matches keywords.

Ulpian allows users to input complex, real-world situations, such as "We are a hospitality company, want to open a pastry shop and cafe, can a person with a pet enter?" rather than forcing keyword searches. The true utility of AI in specialized fields emerges when it can interpret natural language questions, allowing users to describe their problems in context, rather than translating them into a machine's terms.

Ulpian's core promise is to "transparently show you all the relevant provisions," because, as Sandra emphasizes, lawyers "must know where it's prescribed." In domains where accountability is paramount, an AI's output is only as valuable as its ability to cite its sources and reveal the exact origin of its information, making its reasoning verifiable.

The flexibility of language models means users can ask Ulpian questions in any language and receive answers in the same language. For businesses operating across diverse regulatory landscapes, this multilingual capability is not merely a convenience, but a core requirement for frictionless cross-border navigation.

Explore further

ScenariosBusiness challenges CardsEducational cards TimelineTopic evolution