Q1. Could you start by giving us a brief overview of your professional background, particularly focusing on your expertise in the industry?

For the past twenty years, I’ve had the privilege of witnessing the evolution of enterprise technology up close—watching as we moved from simply gathering data, to generating insights, and now to taking meaningful action. I’ve seen analytics mature from basic dashboards to advanced machine learning models, and now to the cutting edge of AI powered by transformer architectures. Every stage has held the promise of change, but for the first time, we’re truly seeing that transformation come to life.

Throughout my career, I’ve always believed that the real value for enterprises lies where data, technology, and AI intersect. I’ve had the chance to put that belief to the test in fields like aviation, automotive, consumer products, and healthcare—whether that meant building pricing platforms for a major jet engine maker or reimagining aftermarket services for a leading healthcare company. No matter the industry, I’ve seen that real transformation is less about the technology itself and more about the value it helps unlock.

Now, as Global Delivery Head at Newpage Solutions, I work with teams around the world to develop AI-driven solutions for healthcare and life sciences. My focus is on growing a strong group of engineers who partner closely with clients and deliver real results—not just presentations. Even after twenty years, I’m convinced we’re still at the very beginning of what AI can do for businesses. For the first time, the technology truly matches the ambition.

Q2. How is enterprise demand for AI solutions changing as organizations move from GenAI experimentation toward production-scale deployments?

It’s a bit of a paradox: instead of slowing down, experimentation has taken off. AI is now accessible to so many that domain experts themselves are building prototypes. Pilot projects are everywhere—78% of enterprises have at least one underway. But the real hurdle isn’t building prototypes anymore; it’s figuring out which ones are worth taking further. Which experiment deserves to become an MVP? Which MVP is ready to scale? That’s where most of the value slips through the cracks: according to MIT, 95% of GenAI pilots fail to deliver measurable ROI, and just 14% of enterprises have managed to scale an agent organization-wide.

Scaling up has always been tough—and large language models have made it even tougher. Now, we’re building on top of probabilistic systems, which means we need new ways of working: things like evaluation processes, guardrails, observability, and strong data governance. The technology is up to the task; it’s the organizational and operational challenges that need solving.

We’re at a turning point. Leaders are starting to move beyond simply adding AI to existing workflows—they’re embracing GenAI that creates AI-native business processes from scratch. Companies aren’t interested in launching more pilots; they want partners who can help them get across the finish line into production. That’s why there’s growing demand for embedded, forward-deployed teams across industries. The era of experimentation gave us a thousand flowers; now, it’s about choosing which ones to nurture—and building the right environment for them to thrive.

Q3. How are concerns around AI governance, security, data privacy and regulatory compliance influencing the way enterprises are approaching AI adoption?

At a fundamental level—and I think this is a good thing—governance has evolved. It’s no longer just the department that says “no”; it’s become the foundation for building trust. These days, the organizations succeeding with AI aren’t just racing ahead—they’re doing so with strong controls in place.
Regulations are changing fast, and everyone is trying to keep up. The EU AI Act, with its new requirements for transparency, enforcement, and penalties, comes into effect in August 2026. And because of its global reach, any company that serves European customers will need to comply. In life sciences—where I spend most of my time—regulators like the FDA and EMA are also getting specific about what it means to use AI in validated, GxP-approved environments. Across the board, the message is clear: if you can’t show control, you shouldn’t deploy.

I see three big changes happening. First, governance is shifting earlier in the process—companies are now building in evaluation, auditability, and human oversight from the very beginning, instead of adding them on just before launch. Second, concerns about data privacy are actually shaping the way organizations design their models: the move toward open-weight models and private deployments is just as much about keeping data safe as it is about managing costs. Third—and this one often flies under the radar—AI agents have created a whole new security challenge: managing non-human identities. When bots have credentials, make API calls, and interact across systems, your security risks aren’t limited to your employees anymore. Most companies still don’t have the right identity and access controls in place for a workforce that includes both people and machines.

Here’s an unexpected perspective: in highly regulated industries, compliance isn’t just a cost of doing business—it’s becoming a real competitive advantage. The companies that make governance part of their DNA—with validated pipelines, traceability, and ongoing monitoring—will be able to use AI in critical areas that others can’t reach. Right now, trust is growing even faster than the technology itself.

Q4. As AI agents increasingly interact with multiple enterprise systems, is orchestration and integration becoming a bigger competitive advantage than the underlying model itself?

It really depends on your perspective. For the leading-edge labs, the answer is more and more ‘yes’—the models themselves are starting to look similar in what they can do, so the real difference comes from how you guide, evaluate, and manage those models. The competitive edge is moving away from just having the smartest model to how effectively you can use and control it.

But for enterprises, I’d look at it another way: orchestration isn’t just a strength—it’s also where the risks show up. Every new system an agent interacts with opens up more ways things can go wrong. In tightly regulated industries like life sciences and financial services, those mistakes can have serious consequences—compliance issues, patient safety, or financial risks. If an agent makes a mistake in a demo, it’s just a glitch; if it happens in a real, regulated process, it could be a major problem.

And there’s another factor: big organizations are naturally cautious—which makes sense—and that means they’re often slow to hand over entire workflows to automation. They might automate a single step, but they’re hesitant to let the system take over the whole process.

That’s where AI-first startups really shine. Their advantage isn’t having a better model—everyone has access to the same technology. What sets them apart is speed: they can experiment quickly, learn from failure, and design their business processes around the agent, instead of trying to fit new technology into old systems. In the years ahead, the real winners won’t just have the smartest models—they’ll be the ones who can combine the agility of a startup with the discipline and safeguards of a large enterprise. Finding that balance is the big engineering challenge for the next five years.

Q5. What is one emerging trend in enterprise AI that investors should watch closely, and what impact could it have on the market?

If I had to sum it up in one phrase, it would be vertical integration. No matter if you’re a cutting-edge research lab, an established company, or a deep-tech startup, the real advantage today isn’t just having the best model—it’s about owning the entire value chain that surrounds it.

Take a look at what’s happening in the market. OpenAI, for example, is working on its own custom chips to make their operations more efficient—because if inference is your main cost, you want to control your own margins, not just rent them. SpaceX and X are going even further, building everything from chips and space-based data centers to collecting their own data, managing model routing, and even creating their own code hub. This isn’t just a product strategy—it’s an ambitious move to own every layer, from hardware all the way to the developer’s desktop.

Enterprises are taking a similar approach, but from the opposite direction. The most forward-thinking companies are working to reduce their reliance on outside labs by building their own models—often using open-weight foundations like GLM and Kimi, and wrapping them in open-source tools. The logic is straightforward: if AI is at the heart of your business, you simply can’t let someone else control the technology or pricing that powers it.

For investors, this shift is significant. The days of simple API wrappers are coming to a close. The real value is moving to those who control the most layers of the stack—compute, data, models, orchestration, and distribution. Expect to see a wave of consolidation: companies focused on just one thin layer will struggle, while investment will pour into those building true depth. In the last big technology wave, the winners owned the ecosystem. This time, the leaders will own the whole value chain.