That is the question at the heart of a recent conversation between Don Finley and Rahul Parekh, a venture capitalist at 2150, a fund dedicated to investing in AI for the physical world. Their discussion covers a remarkable breadth of ground — from cement plants to electric trucks, from the future of work to parenting in an age of intelligent machines — and offers a refreshingly grounded perspective on where AI is headed next.
From Trading Floors to Factory Floors
Parekh’s path to this niche is unconventional. He began his career as a derivatives trader at Goldman Sachs before moving into venture capital, then leaving to build and operate his own company, and eventually returning to the VC world to co-found 2150. That detour into operating, he argues, was not a sidestep but an essential education.
Being a founder gave Parekh firsthand experience with the emotional rollercoaster of building a business — the firefighting, the constant toggling between strategic vision and being deep in the weeds, the highs of customer love and the lows of everything breaking at once. That empathy, he says, is something you simply cannot develop from the investor side of the table. It shapes how he evaluates founders and how he thinks about the gap between a great product and a successful business — a gap that is almost always bridged by human factors rather than technical ones.
The Physical World as AI’s Next Frontier
2150 focuses on five sectors that together represent the backbone of modern civilization: energy, buildings, factories, mobility, and infrastructure. These are domains where technology adoption has historically been glacial. Cement plants, steel mills, construction sites — these are industries where core processes have remained largely unchanged for decades. The barrier has always been that implementing new technology in physical systems is expensive, slow, and requires massive operator involvement.
AI is changing that equation. The key insight is that physical systems are now generating enormous amounts of real-time data, and AI can use that data to optimize operations without requiring a wholesale overhaul of existing infrastructure. Parekh points to a striking example from the cement industry. A single cement plant optimized with AI-driven self-learning control systems can save over a million pounds annually, reduce CO2 emissions by thousands of tons, and lower output variability — all measurably and all without replacing the plant itself.
This is a fundamentally different value proposition from the AI we have seen in the digital realm. It is not about generating text or answering queries. It is about running complex industrial processes more efficiently than human operators could manage alone, learning in real time, and adjusting against operator-set targets. And critically, it is measurable — you can quantify the savings, the emissions reductions, and the productivity gains in a way that has been elusive for much of the AI hype cycle.
The Compounding Loop: Design, Manufacturing, and the Real World
Another powerful example comes from the mobility sector. Parekh’s fund recently invested in a company called Superlite, which has built an electric vehicle designed specifically for middle-mile logistics — the trucks that move goods between distribution centers. The vehicle, called the OV1, is a seven-and-a-half-ton truck that carries twice as many pallets as comparable internal combustion or electric vehicles.
What makes this more than just a better truck is the data loop. Every vehicle is instrumented and feeds performance data back into an AI-powered digital twin of the production floor. This creates a compounding cycle: real-world performance informs design, design informs manufacturing, manufacturing informs the next generation of vehicles, and the whole system gets smarter over time. It is a closed loop between the physical and the digital that would have been impossible without AI, and it represents a new paradigm for how products are designed, built, and improved.
The broader implication is that this kind of AI-enabled logistics could ripple across the entire e-commerce ecosystem, making supply chains more efficient, reducing emissions, and creating new economic opportunities that simply were not viable before.
The Jobs Question: Displacement, Creation, and the Uncomfortable Transition
No conversation about AI is complete without addressing the elephant in the room: what happens to jobs?
Parekh is candid that some roles will disappear and some teams will shrink. If one person with AI can do what three people did before, that is a reality that cannot be ignored. But he is also optimistic that AI will make things economically possible that were not before, and that new categories of work will emerge as a result.
The distinction he draws is between types of work. Knowledge work — gathering information, analyzing data, producing first drafts, writing code, making presentations — is changing faster than most people expected. AI is getting remarkably good at these tasks. But work that involves ambiguity, responsibility, relationships, and physical complexity is much harder to automate. In those areas, Parekh argues, human roles will not just survive but actually expand, because AI will enable people to unlock outcomes that were previously out of reach.
Don Finley adds a complementary perspective from his own experience in logistics. He shares a telling story about deploying IoT sensors in vending machines to optimize delivery routes. The technology worked — it could pick the right machines in 10 minutes instead of four hours. But a delivery driver who had spent years personally selecting machines refused to trust the system, even when shown results from other markets. It was not malice or ignorance driving his resistance. It was care — care for his customers, care for the organization, and a deep personal investment in doing the job the way he had always done it.
That story captures something essential about the AI revolution that is often overlooked: the technology is the easy part. Changing behavior is the hard part.
Advice for the Next Generation
When the conversation turns to what advice they would give young people growing up in an AI-saturated world, both guests converge on a similar theme: know yourself.
Finley argues that the most important thing anyone can do — especially young people — is to develop a deep understanding of what genuinely interests and motivates them. AI will handle more and more of the mechanical aspects of work, which means the human differentiators become relationships, strategy, creativity, and judgment. He draws an analogy to calculators: you learn to use the tool, but you do not delegate your critical thinking to it.
Parekh, speaking as a father of two young children, adds that the next generation should not try to compete with AI at the things AI will inevitably do better. Instead, they should learn to use AI extremely well while building the skills that are harder to commoditize — selling, communicating, understanding people, developing judgment, and staying curious. He describes a practice he uses with his own kids: tackling difficult problems together, working through how they would solve them, and then comparing their approach to how AI would solve the same problem, and finally asking another person for their perspective. It is a way of teaching children that the real world is messier and more nuanced than any single approach can capture.
What Being a Parent Teaches You About Investing
One of the most compelling threads in the conversation is Parekh’s reflection on how fatherhood has shaped his investment thesis. Kids, he says, teach you that reality is messy. Customers do not behave the way you expect. People do not adopt products simply because they are technically better. In the industries 2150 invests in — factories, energy systems, construction — you cannot walk in and win with a better algorithm. You have to understand why someone would change a process they have used for twenty years, and the barriers are almost always human: fear of losing responsibility, distrust of the system, worry about being blamed if something goes wrong.
This is why execution and adoption matter more than technical superiority, and why the most successful founders are not necessarily the ones with the best technology but the ones who can build relationships, communicate effectively, and create the distribution channels that turn a clever product into a real business. Technology can be the easy bit. Changing behavior is the hardest part.
The Bottom Line
As AI continues to evolve, the most interesting applications may not be the ones that generate content on a screen but the ones that quietly optimize the physical systems we all depend on — the cement in our buildings, the trucks on our highways, the energy powering our cities. The technology is becoming more capable than most people imagine, but the real question is not what AI can do. It is what we choose to do with it, and whether we can bring people along for the ride.
The conversation between Don Finley and Rahul Parekh is a reminder that the future of AI is not just a technical story. It is a human one.
