Almost every meaningful event that occurs in the world has happened because humans connected. Every company that got started, every hire that changed someone's career, every customer relationship that turned into something bigger, every breakthrough idea that built upon someone else's work: it all has come down to humans connecting with other humans. In an unromantic sense, the progress of humanity has been built on real human relationships. Is this all about to change because of AI?
In February 2026, an article titled "Something Big Is Happening" went viral. It was one of the first moments when I witnessed people outside of tech wake up to what those of us inside it have been seeing for a while.
Since then, the conversation has only gotten louder. In September, a researcher's resignation from one of the leading AI labs set off a wave of headlines about AI and the extinction of humanity. A few days later, Anthropic's CEO, Dario Amodei, published "We Must Pace the Frontier," arguing that the industry should deliberately slow down so that safety can keep up. And then the President signed an executive order directing federal agencies to stop saying "Artificial Intelligence" and start saying "Super Intelligence."
Clearly, AI (or should I say SI?) is now on everyone's mind.
So I'd like to share where I stand. I am neither an AI doomer nor an AI skeptic. I'm a technologist, with a Ph.D. in computer science. I co-founded Carta, and as an early-stage investor I backed companies like Twilio and Lyft before most people had heard of them. I use AI every day and I've built agents that talk to my other agents. By temperament and by training, I am about as excited about this moment as a person can be.
Given current headlines, most people think AI happened overnight. It didn't. AI has been a field of serious research for decades. In the 1990s, as a student at Carnegie Mellon, I worked directly under Raj Reddy, one of the pioneers of artificial intelligence and a winner of the Turing Award, the closest thing computer science has to a Nobel Prize. Later, at Stanford, I worked under Prof. Terry Winograd, whose 1971 dissertation work produced one of the first programs that let a computer carry on a conversation in plain English, decades before any of us typed a question into ChatGPT.
What changed recently is that decades of advances in computing and AI research have finally converged. Fei-Fei Li's ImageNet project gave machines millions of labeled images to learn from, AlexNet, built by Alex Krizhevsky, Ilya Sutskever, and Geoffrey Hinton, showed what neural networks could do with them, and the chips NVIDIA designed to render graphics turned out to be exactly what was needed to train AI models. Each breakthrough propelled the next, and the outcomes compounded. Today, silicon atoms strung together in the form of microchips can simulate intelligence at the level of some of the smartest people on the planet. And not in just one field, but in many fields, all at once. So yes, something big is happening, and happening fast.
As a computer scientist, I find the shift profound. For my entire career, everything a computer did had to be deterministic and clearly defined. You gave it the same input, and it gave you the same output, every time. Now we can ask a model to exercise judgment. We have moved from deterministic computing to judgment-based computing. We now have intelligence as a primitive, the computer science term for a basic building block that everything else is built on, and we can use it to build entirely new kinds of systems.
I'm not watching this from the sidelines. I started with ChatGPT like many of us did, then moved to Gemini, then added Claude, and now Grok and Grok Bot. My agents watch the messages coming into my inbox, flag the ones I need to review, and in some cases act on them on their own.
My graduate focus at Stanford was Human-Computer Interaction, the part of computer science that studies how people and technology work together. This field is even more important today, especially as we figure out how humans will interact and engage with advanced AI, and how people and technology can work together in a way that is symbiotic (hopefully). The computer scientist J.C.R. Licklider imagined exactly this kind of "Man-Computer Symbiosis" back in 1960. We appear to finally be building it. And while the geek in me is thrilled about LLMs, agentic intelligence, and self-driving cars, I believe it's the interactions among people that are about to matter more than ever before.
I want to be clear about one thing up front. This isn't an essay about jobs. Like every major technological transition, from horses to cars, or from building by hand to using heavy machinery, AI will end some jobs and create new ones, and how we navigate that as a society is a serious question that deserves its own conversation. My argument is about something else: what happens to the relationships between people.
Here is my thesis. As AI takes on more of the mundane tasks that people perform, it will create both the capacity and the need for more human-to-human connection. The more of my work my agents take on, the clearer it becomes that the part they can't do is the part that matters most. When everyone has the same AI models and the same AI tools, the scarce thing isn't another agent. It's a real relationship with another person.
Human connection is going to become more important in the future, not less. That is what gives me hope that we are fundamentally headed for a better future.
Our need for connection with other people is deeply rooted in our evolutionary biology. It isn't a preference or a habit. It is foundational. Humans evolved over hundreds of thousands of years as social creatures, and a few years of chatbots will not rewrite this.
Human connection is at the core of how people communicate, establish trust and credibility, and decide whether to work with each other. This has been the cornerstone of business for centuries. Think about the big decisions you've made in your working life: the job you took, the person you hired, the partner you chose, the vendor you bet a project on. Did you make those decisions purely on the information in front of you? Or at some point, did they come down to a person you knew, or a trusted person who vouched for them? While information has always been part of the decision, trust is what lets you act on it.
Today, AI makes information abundant and nearly free. It can draft, summarize, and analyze faster than any of us can think. But it can't look you in the eye across the table, shake your hand to seal an agreement, or put a hand on your shoulder and tell you it's going to be okay. It can't share a meal with you or get on a plane to be there in person. Those moments carry weight because a person chose to spend their time, their presence, and their reputation on you.
In this AI world where any company can quickly generate a polished website, a stack of white papers, and a thousand personalized emails in an afternoon, trust in a company will mostly be established by its people. The connections, interactions, and relationships these people have are how a company's brand will be genuinely experienced. Companies invest heavily in their people: hiring, training, developing, and managing them. They also invest considerably in their brand. Yet those two investments rarely meet. The opportunity ahead is ensuring that the company's brand is experienced through its people. A prospect forms an impression of a company from the person who shows up to the meeting, the email that lands in their inbox, and the follow-up that arrives three weeks later. The authenticity of an actual connection is the part that is hardest to fake.
Here's the part I find most interesting. AI won't just leave room for human connection. It will raise the bar for it.
For years, it has been easy to Google someone or look them up on LinkedIn before you meet them. Now we don't even need to do that ourselves. AI can do it for us and prepare a brief on who we're meeting, what they care about, and what to ask. Every conversation can be recorded, transcribed, and analyzed for key points and action items before you've walked back to your car. Every follow-up email can be drafted and sent before you get back to your desk.
When everyone shows up to the meeting with the same perfect brief and leaves with the same perfect summary, the information stops being an advantage. It becomes table stakes. What's left to separate one person from another is the real relationship itself: whether you actually listened, whether you remembered what mattered to them, and whether they believe you'll do what you said you would.
Meanwhile, the volume of everything else is exploding. AI-written outreach, bot-generated replies, and content produced at a scale no human could match are flooding every inbox and every feed. Reaching a thousand people has never been easier. Our error is that we've started mistaking that digital exhaust for connection. A follower is not a relationship. A notification is not a conversation. We measure everything from whether someone opened an email, how long they looked at an ad, and which link they clicked. However, we have almost never measured whether two people actually connected.
As the noise goes up, real signal becomes even more valuable. Truly connecting with one person is as rare as it has ever been. I fully expect the bar for our human relationships to go up, and the people and companies who clear it will stand out more than they ever have.
There are two serious objections to all of this, and I don't want to wave them away.
The first is that agents will soon be doing the deciding. Your agent will negotiate with my agent, compare a hundred vendors in seconds, and make the purchase. If machines are transacting with machines, why would human relationships matter at all?
I don't think anyone has fully worked this out yet. But my instinct is that even when agents make decisions and transact on our behalf, human relationships will still influence those decisions. Agents act on preferences, and on lists of trusted sources and approved partners, that people set. Relationships shape those lists. And when something goes wrong, as it eventually does, nobody wants to escalate to an agent. They want a person who is accountable.
The second objection is that people are already forming relationships with AI itself. Many people talk to AI companions every day, and some find real comfort in them. If a chatbot can be a friend, maybe we'll need each other less.
I take that seriously, too. An AI companion can be patient, available, and kind. But a relationship with another person carries two things an AI can't offer: reciprocity and stakes. A person can be appreciative of you or disappointed in you. They can choose to show up for you when it costs them something, or not show up when it costs them nothing. That is what gives trust between people its value and meaning.
I have always believed that magic happens when humans connect. It doesn't always feel like magic in the moment. Sometimes it's a 30-second exchange at a conference or a follow-up email three weeks later. But those small moments compound, and over time they shape entire careers and companies.
I believe this so strongly that while many of my colleagues are out building better AI, I have spent the last eight years building a company that helps people connect better and stay connected.
I'm going to keep using AI every day. I'm going to keep building with it, investing in it, and marveling at it. And as it continues to take on more of your own work, I'd ask you to do one thing: spend the time it gives back on people. Make the phone call. Follow up. Show up in person when you can. Remember what truly matters to the person sitting across that table.
In a world with AI, human connection matters more. Every good thing still starts with two people in a room.
Dr. Manu Kumar is the co-founder and CEO of HiHello, a Professional Presence and Lead Capture platform that helps companies take charge of how their people show up and connect. He co-founded Carta in 2012 and is the founder of K9 Ventures, an early-stage fund that was an early investor in Lyft, Twilio, Lucidchart, Auth0, Everlaw, and Aurora. He holds a Ph.D. in computer science from Stanford, where he focused on Human-Computer Interaction.
