Artificial intelligence is no longer a side initiative. It’s becoming the operating system of the enterprise. Dell Technologies has used AI to transform how its teams work, enabling it to decouple revenue and cost while improving customer outcomes. Mercedes-Benz is redefining personalization through AI-powered conversational interfaces and software-defined vehicles. AMD is building AI fluency company-wide through mandatory training and responsible governance. And Liquid AI is pushing intelligence beyond centralized data centers and onto the edge—into vehicles, devices, factories, and industrial systems. The question these leaders are answering is no longer whether AI will transform the enterprise, but how fast and how deeply.
McKinsey senior partner Jeff Hart sat down with Magnus Östberg, chief software officer at Mercedes-Benz; Ramin Hasani, cofounder and CEO of Liquid AI; Ruth Cotter, senior vice president and chief administration officer at AMD; and Sam Burd, chief strategy officer at Dell Technologies to discuss what becoming AI-native requires and how AI is reshaping growth, personalization, and the workforce. The following is an edited transcript of their conversation.
AI as a growth imperative
Jeff Hart, McKinsey: The question now is not whether AI will transform organizations but how leaders will use it to reshape how they engage with customers and employees. What is the unlock that gets us both speed and scale of deployment?
Ramin Hasani, Liquid AI: Efficiency is critical. We have to work on the efficiency of models to satisfy the exponential demand AI is creating. The approach we’re taking is pushing intelligent systems outside of data centers—leveraging processors already embedded in cars, mobile phones, laptops, airplanes, IoT [Internet of Things] devices, and factories. If you can bring intelligence directly onto those devices, you create a major unlock. Billions of devices already exist. What’s needed now is the technology breakthrough to make AI on the edge fully viable.
Jeff Hart: Operationalizing AI—actually scaling and deploying it—is still catching up to the ambition. How do companies close that gap?
Sam Burd, Dell Technologies: Through our own efforts to modernize our company with AI, we’ve learned that this has to be led from the top. The real inflection point for us came about three years ago, when we began this journey. Michael Dell challenged the senior leadership team with a stark premise: Five years from now, we’re going to have a new competitor that is faster, more innovative, and more successful at a lower cost, and they’re going to put us out of business. And the only way to prevent that is to become that company.
That reframed the conversation. This wasn’t about side projects or isolated proofs of concept. We started from our core differentiators—our products, our services, our supply chain, our go-to-market organization—and committed to using AI to make each of them stronger.
Over the past two years, that focus has allowed us to decouple revenue and cost while delivering better outcomes for customers at the same time. When people ask us how we are growing revenue when our costs are going down, the answer is clear: We’re using AI to fundamentally improve how we operate. And importantly, this is not just a technology journey; it’s a business transformation. For it to be successful, your leadership needs to be 100 percent on board, aligned on desired outcomes, executing on one clear strategy, and inspiring the same in their team members.
Jeff Hart: What are companies most commonly underestimating?
Sam Burd: That fundamentally this is a people transformation. Technology enables you to operate differently, but leadership commitment, strategy, and employee adoption are what will determine whether you actually capture value. If I see an IT leader who says they alone have been charged with implementing AI across the company, then I’ve got some bad news for them. If technology teams are the only ones on the journey, you will not win.
Reimagining customer experience
Jeff Hart: Magnus, Mercedes-Benz sits at the intersection of software, hardware, and customer experience. How important is software becoming to the future of the brand?
Magnus Östberg, Mercedes-Benz: It’s becoming a central focus. Mercedes-Benz is also an aspirational brand with strong generational loyalty, so we’ve spent considerable time thinking about how AI and software amplify both the brand and the business, not just automate it.
The internal transformation is significant. We’ve fundamentally decoupled hardware from software so that software can operate on its own life cycle and at its own pace. That transformation enabled what we call the Mercedes-Benz Operating System—a software-defined vehicle platform. But at the end of the day, two things matter: Can the company generate growth and return on investment, and do customers love what we’ve built?
One of the things that made the most news was being the first to integrate ChatGPT into our vehicles. People initially asked, “Who wants ChatGPT in a Mercedes-Benz?” It turns out customers love it. What’s interesting is what they actually do with it: They talk to the car, ask for advice, engage conversationally. The car becomes a private, luxurious space where customers can be productive or simply interact naturally.
Jeff Hart: How much of the future strategy is about making those experiences more personalized?
Magnus Östberg: Entirely. One of the most significant advantages of software and AI is that I no longer need to cluster customers by age or region. I can individualize the experience. The digital cockpit becomes a canvas, and increasingly, that experience won’t be statically designed by us. It will be generated dynamically using AI models based on each customer’s preferences, patterns, and behaviors.
At the same time, trust becomes critical. Our brand promise extends beyond physical safety to digital safety. Customers need confidence that what they say and do inside the vehicle remains private and secure. That’s why we’re investing heavily in moving from cloud-based AI models toward on-device intelligence, where data never leaves the car.
Ramin Hasani: That’s exactly where the industry is heading. We need a balance between cloud models and embedded models running directly on devices. Intelligence increasingly needs to happen where the interaction occurs; physical AI and edge AI are becoming indispensable for that.
Building AI fluency at scale
Jeff Hart: Ruth, there’s a clear tension between speed, quality, governance, and risk. How are you balancing that at AMD?
Ruth Cotter, AMD: It starts with leadership commitment. Tone from the top really matters. One message that has resonated internally is simple: Don’t get left behind. Whether you’re thinking about your own career or your company’s competitive position, someone will outcompete you at a speed the world has never seen before.
At the same time, we have to be responsible. If you over-index on process and governance, you slow people down to the point where adoption stalls. So we’ve focused on establishing responsible AI principles early in the workflow, rather than layering on heavy processes later. We have a responsible AI group internally, but ownership still sits within the business teams themselves, so they can move quickly. The goal is clear guardrails without bureaucracy for its own sake.
Jeff Hart: You’ve also placed a major emphasis on training across the organization.
Ruth Cotter: Yes. We’re mandating AI training for all employees, because we want everyone to reach a baseline level of proficiency. It’s about creating consistency, but it’s equally about signaling seriousness—making clear that AI fluency is not optional and that it matters across every function.
Earlier in our journey, we gave employees about nine months of freedom to experiment, simply to build familiarity and confidence. That made a significant difference. People became much more willing to engage because they understood the tools. We’re also hiring more interns than ever and pairing them with experienced engineers. Newer employees often come in highly proficient with AI tools, while veteran engineers bring deep technical expertise. Combining those perspectives has accelerated curiosity and adoption across teams.
Sam Burd: We’re seeing something similar. Across the broader landscape, the top 10 percent of AI users are disproportionately consuming tokens and capabilities. There’s a real divide emerging between people who are deeply engaging with AI and those who aren’t, which is why leadership participation matters so much. We need leaders to show people the art of the possible with AI in their day-to-day work, give them room to experiment, and also set clear boundaries and expectations. Our sales teams, developers, and operations teams are expected to use these tools, but they’re also encouraged to help shape how we use them. That’s how we’ll continue creating better outcomes for ourselves and for customers.
Scaling from experimentation to enterprise value
Jeff Hart: Many organizations remain stuck in proof-of-concept mode. What separates the companies that are actually scaling?
Sam Burd: One of the biggest mistakes companies make is treating proof of concept as the finish line. Then what happens is you launch a tool or capability, people engage with it briefly, and then everyone moves on. What we’ve learned is that iteration matters enormously.
If we had launched some of our early tools and stopped there, people would have said, “That’s interesting,” and returned to old ways of working. Instead, we committed to continuously improving the experience—better data, constant refinement, leadership engagement, user feedback. You have to be fully committed. Otherwise, you end up halfway in, wondering why the ROI never materializes.
Ruth Cotter: One challenge we’re navigating now is distraction from too many tools. Adoption is good, but going too broad creates fragmentation instead of consistency. That’s why a more deliberate, top-down strategy matters. You need longevity and repeatability in what you’re building.
Data is also foundational. Early on, we made the decision to build our own internal data system, because we knew fragmented data would become a major bottleneck. There was frustration at first, because the ROI wasn’t immediate, but it created the foundation we needed to move much faster later.
Jeff Hart: Ramin, as someone building foundation models while also thinking about where AI is headed, how should companies invest for both today and tomorrow?
Ramin Hasani: One of the biggest opportunities is building for yourself. AI gives organizations tremendous internal capability. But companies need to actually empower employees to build and experiment with it. We may soon see one-person unicorns because these systems so dramatically amplify individual capability. Imagine what becomes possible inside enterprises with tens of thousands of employees.
But organizations need to invest internally to unlock that potential. They need to continuously upskill and stay close to the state of the art, because innovation is moving incredibly fast. When I look ahead, I think we’re moving toward the automation of automation—a future where organizations can eventually automate portions of their own R&D and where innovation itself is something AI helps drive.
