The first panel of Forbes Business Bridges 2025 opened with a forward-looking discussion on how artificial intelligence, automation, and digital ecosystems are redefining competitiveness and innovation across industries.
Moderated by Mihai Ivașcu — Founder and Board Member of Modex, Co-Founder of .Lumen and Aeronews, Board Member of Rayscape.ai, and Managing Partner at I64 Capital — the conversation brought together a powerful lineup of innovators: Adrian Gociu (Gociu Lawyers, LoContemporary.com), Marian Anghelache (Global Vision), Constantin Gorgan (CG Tech Labs, Giles AI), and Andrada Morar (UiPath).
Below are the main insights shared by the speakers:
ADRIAN GOCIU – Lawyer, Founder & Managing Partner Gociu Lawyers, Founder LoContemporary.com
Right now, the EU, the U.S., and even China are converging in interesting ways, though with different styles. In the European Union, the approach is more prescriptive: regulate more and define what happens at each step. In the United States, the legal framework leans on fair use debates and the common-law tradition. These differences impact how AI is regulated and how we think about it in each jurisdiction.
Here in the U.S., we don’t yet have many finalized common-law cases that clarify liability in AI. Because of that uncertainty, some funds are cautious about investing in AI-made products or software. A practitioner I spoke with in Silicon Valley last week told me it’s becoming extremely difficult—almost impossible—to trace which parts of code were AI-generated just from the output.
So the challenge isn’t only how we regulate AI, but also how we prove AI’s involvement and provenance. We generally know that much of today’s AI has been trained on copyrighted data. Often the outputs may not infringe copyright, but there are edge cases. We’ve also seen high-profile lawsuits—for example, involving Anthropic—with very large damages being sought over training data. Some of those can be proven; in general, it’s still hard.
Mihai Ivașcu: Let me give you a concrete example. Say we’re navigating a blind person around town and the system says, “You’re clear to walk,” but they aren’t—and a car hits them. Or consider a medical AI that claims 99.8% accuracy in detecting cancer on CT scans, but returns a wrong result. Who is ultimately liable for AI’s outcomes?
Gociu: It’s a great question. At every conference, liability is the topic. It’s probably too early to define exact liability across the board. The law will evolve—especially in the U.S. under common law.
At this moment, the end user often bears primary responsibility, and most AI tools include extensive disclaimers. That said, I expect this to change, given the scale of AI businesses and the money involved. If providers faced no liability, that would be problematic. Also, insurers—particularly in the U.S.—are starting to introduce new policy rules (think autonomous features in cars, for example). How insurers decide to cover losses will strongly influence how liability is allocated.
MARIAN ANGHELACHE – Head of Technology and Innovation Global Vision
Last year we talked about prompt engineering and AI felt like a novelty. This year the focus is on context engineering—and that’s becoming the new standard. You’re no longer just prompting or asking questions; the AI can execute business tasks if you put the right context in front of it.
Data alone means little unless it’s structured and placed in the right context for that company. That’s why we built our platform: a contextual AI, a custom enterprise platform. Think of it as the gear shift between AI as a technology and people as the pilots. But you can’t drive it if you don’t shape the context correctly.
That “right shape” is context. When we deploy AI in an organization, AI doesn’t transform the company by itself—people do. People with the right context and the right tools deliver the transformation.
(…) We build custom AI agents for each industry. We’re focused on real estate, finance, and investment, where the heavy lifting is significant. In real estate, for example, data lives in silos and must be curated and contextualized.
Say you have a pipeline of projects: you need to validate each one—land ownership, permit checklists, whether permits match the location, environmental risk, potential compliance or due-diligence issues. We bring all of these into one context and generate a checklist for the stakeholder.
The stakeholder doesn’t have to write prompts—we provide the agenda and do the heavy lifting. Humans make the final call; they remain in charge. What matters is giving the AI the right coordinates. Otherwise, it can hallucinate and lead to bad decisions.
CONSTANTIN GORGAN – Founder @ CG Tech Labs & Co-Founder and CTO @ Giles AI
The biggest difference isn’t about ideas—it’s about velocity and capital density. In Silicon Valley, you’ll hear people discussing startups and moving millions at every coffee shop. In Europe, that’s not the case. Early-stage companies often find it harder to secure the capital they need at the starting line, at least until they show real traction.
I’ve seen companies in Silicon Valley reach a Series A in six weeks—that’s unheard of in Europe. One key takeaway is speed: my research shows that, on average in Europe, the same round takes 8 to 12 months.
Another important point: it can be smart to leverage Europe as your launch environment. There are advantages—lower costs, strong talent pools, and easier access to skilled people who might be harder (or more expensive) to hire in the U.S. ”
ANDRADA MORAR – VP Customer Experience & Global Partners UiPath
It’s interesting — we heard earlier today how much investment has been made, and continues to be made, in AI. We’re seeing that both from the funding side and from the enterprise side. Yet, there’s an unfortunate statistic: about 95% of AI projects fail. Why? If we look at our own journey — from RPA to Generative AI — the answer is actually pretty simple. Many companies focus on the “sexier” part of AI, which often means building yet another chatbot. But that doesn’t necessarily create real impact, especially at the enterprise level. At UiPath, we started with RPA — Robotic Process Automation, and we continue to believe it’s the healthiest foundation for GenAI. Think of it this way: RPA is excellent at role-based automation. If you tell a robot to go from point A to point B, it will go exactly from A to B.
If you tell an AI system to do the same, it might also go from A to B — but not necessarily by taking the route you expect. When we evolved our vision of emulating and augmenting humans, we began to see Generative AI as an incredible opportunity to reimagine how work gets done — not just to automate tasks, as RPA does, but to transform entire business processes end-to-end.
That’s where we see our customers achieving real success. They’re not focusing only on one task or a “cool” part of a process; instead, they have the courage to take on truly hard enterprise challenges — the ones they wouldn’t even touch before, because the technology wasn’t ready.
Those are the companies that make up the successful 5%. And what sets them apart is their approach: they have CEO-level support. They use RPA where RPA works best, AI where AI works best, and humans where humans are irreplaceable. And they bring it all together under a framework with clear governance and auditable systems.
That’s the model we believe in — where technology doesn’t replace people but empowers them, and where automation and AI work hand in hand to truly reshape how enterprises operate.
Moderator: MIHAI IVASCU – Founder and Board Member of Modex – Co-Founder of .Lumen – Co-Founder of Aeronews – Board Member of Rayscape.ai – Managing Partner I64 Capital
It’s a great time for AI. I want to set the stage for what is probably one of today’s most interesting discussions—not only in business, but in the world and in our daily lives. AI has moved beyond the hype. It’s not just a buzzword; it’s an active technology delivering real results. In our ventures, for example, with AI we’ve helped blind people walk. With AI on the RaceKeep project, we ran a test in under 45 seconds. ModX used AI for FIFA ticketing. These technologies are already working behind the scenes and achieving great outcomes. That’s why so many top developers—ambitious people who want this technology used well—are drawn to it. More than that, this is a paradigm shift. Consider this:
Ninety percent of the world’s data was produced in the last 24 months. Of that, only about 1% is being properly used and interpreted. The room for AI-driven growth is enormous. We shouldn’t be afraid of it. We should think carefully, of course, and build the right regulatory frameworks. But what’s better than a great AI tool? An expert using a great AI tool. Instead of resisting adoption, we should embrace it and find ways to use it in our daily lives.
I’m happy to be on stage today with a great panel of speakers, whom I’d like to invite to join me so we can begin the discussion. We want this to be interactive—please contribute and challenge us.