Annie Tsai

I spent two days at HumanX in San Francisco last week, where roughly 6,500 people from nearly 40 countries came together to talk about artificial intelligence. There were 350 speakers, hundreds of demos, and more AI-generated slide decks than I could count. Every major thesis about where this is all going was on full display, and the thing I couldn’t stop thinking about was a random conversation I had on the expo floor with someone whose name I didn’t catch until twenty minutes in.

We were both standing near one of the demo stations, neither of us watching the demo, and she said something about how her company had tried to automate their customer onboarding with an agent and it had gone sideways in a way that taught them more about their product than two years of user research had. I asked a follow-up question. She gave me an answer I wasn’t expecting. I pushed back on part of it. She pushed back on my pushback. We went back and forth for a while, and by the time we exchanged information, I had three ideas I hadn’t walked in with and a connection to someone I genuinely want to talk to again.

Everybody Had AI But Not Everybody Had Something to Say

The interesting thing about 6,500 people building in the same ecosystem being in the same building: the technology itself kind of stops being the differentiator. Within the first few hours, the booths start blending together, and what stands out are these passing conversations where you pick up tidbits of insights here and there until the end of the day you’ve amassed a treasure trove of ideas to vet.

I started paying attention to what those conversations had in common, and it came down to two things. The person had a distinct point of view — a real opinion, shaped by real experience, that they were willing to defend. And they could communicate it in real time, unscripted, and in a way that made me want to keep talking.

That combination is becoming the rarest and most valuable professional skill in the age of AI. We are sick of being talked at, and we want to jump into the ring with you and duke it out, for the both of us to come out better in the end.

The Research Behind Why This Matters

In 1973, a Stanford sociologist named Mark Granovetter published a paper called “The Strength of Weak Ties” that has since accumulated nearly 70,000 citations across business, economics, psychology, and sociology. His finding was counterintuitive and has held up for half a century: your casual acquaintances — the people you meet at conferences, the friend-of-a-friend, the person you end up next to at dinner — are more valuable to your career than your close friends are.

The reason is structural. The stronger the tie between two people, the more their knowledge, networks, and information overlap. Your close colleagues know what you know. Your weak ties connect you to entirely different networks, which means they carry non-redundant information like the things you haven’t heard yet and the perspectives that don’t exist in your bubble.

A 2022 study published in Science looked at more than 20 million people on LinkedIn over five years and confirmed Granovetter’s thesis at massive scale: moderately weak ties, or people who share roughly 10 mutual connections with you, were the most valuable for job mobility. In the end, it’s the in-between people that you meet in hallways and at conferences and through one-degree introductions, that drive the most thought and network expansion.

Here’s the part that connects to what I experienced at HumanX this week: the weak ties don’t activate themselves. They activate through conversation and proactive introduction. Specifically, through the kind of unplanned, serendipitous conversation where two people discover they have something genuinely interesting to say to each other. Google claims that chance encounters among employees were responsible for innovations like Gmail and Street View. Steve Jobs designed Pixar’s offices to force engineers and designers into hallway collisions because he understood that serendipity is the engine of innovation. Research on workplace design has consistently found that at least a third of innovation ideas come from serendipitous conversations that start casually and morph into something bigger.

You run into someone, you say something that surprises them, they say something that surprises you, something new emerges that neither of you would have arrived at alone. That’s the entire process, and it depends on two people who each have something worth saying.

AI Commoditized the Thing Most People Were Offering in Those Conversations

For most of professional history, the thing you brought to a hallway conversation was information. You knew something the other person didn’t, read a report they hadn’t seen, or maybe talked to a customer they hadn’t talked to, or understood a technology they were still learning about. Information asymmetry was so frequently the currency of networking, and having more of it made you more interesting.

AI and the velocity by which people are working in the AI space has collapsed that asymmetry in ways that can make conversations feel so impactful. Everyone at this event essentially had access to the same models, same research, and same synthesized knowledge. Anyone could build a smart-sounding take on any topic in thirty seconds, and the informational advantage that used to make hallway conversations valuable has been compressed into something anyone with a phone can produce.

What AI has not commoditized is interpersonal skill, which is what McKinsey’s research on skill automation specifically identifies as the least automatable category of professional capability. The ability to read a room in real time, to hold a point of view under pressure, to respond to something unexpected with curiosity instead of a rehearsed answer. Their report on human-AI skill partnerships found that capabilities rooted in social and emotional intelligence, including things like negotiation, coaching, and interpersonal conflict resolution, remain uniquely human and are becoming more valuable as the technical layer becomes more accessible.

So the math has changed. If your conference conversation is “here’s what I know about X,” you’re now competing with everyone’s phone. If your conference conversation is “here’s what I believe about X, and here’s what I’ve seen that makes me believe it,” you’re offering something no model can generate. The distinct point of view, earned through experience, sharpened through real human exchange, is the new scarce resource.

You Can Use AI to Sharpen the Blade, But You Need Humans to Use It

I don’t think the answer to “how do I get better at this” is to put your phone away and go talk to people, though you should also do that. AI is genuinely useful for developing the underlying skills, and the research supports it.

Virginia Tech researchers built a tool called CommCoach in early 2026 that simulates difficult workplace conversations and gives users real-time coaching feedback. It helps people notice their habits — how they respond to disagreement, whether they rush to solve instead of listen, how they handle emotional cues — in a low-stakes environment where mistakes are cheap. Skillsoft built CAISY, a conversation simulator specifically for business and leadership skills, and companies like Yoodli have built enterprise roleplay platforms where people can practice pitches, demos, and crucial conversations against an AI that adapts to their tone and arguments.

These tools work, and they work for a specific reason: they let you rehearse the components of good conversation like clarity, perspective-taking, directness, listening in a space where you can fail without consequence. If you’ve never stress-tested your point of view against someone who disagrees with you, doing it with an AI first can be incredibly useful. If you want to practice saying something clearly in sixty seconds, an AI coach can help you get there faster than practicing alone.

The other equally important half of this equation is the unpredictability of human nature which AI can be less helpful with. When you’re standing in a hallway at a conference and someone’s face changes because you said something they disagree with, and you have to decide in the moment whether to soften, hold, or get curious while maintaining your expression… that’s a rep that no simulator can provide. When the conversation goes somewhere you didn’t plan for and you have to think on your feet instead of following a script, that’s where the muscle actually develops. When you hold your point of view and stay genuinely open to being wrong at the same time, in front of a person who can see your face and hear your voice — that’s the version of this skill that differentiates you, and it only develops through practice with other humans.

Use AI to prepare, sharpen your arguments, stress-test your thinking, and rehearse your clarity. Then go have the actual conversations where the outcome isn’t predetermined. The tools will make you sharper, and the people will make you better.

Cultivate Your Hallway People

One more piece of this that I think gets overlooked: the serendipitous conversation can’t happen if you’re not in the room. And being in the room has a specific meaning in the age of AI that it didn’t have five years ago.

When information was scarce and unevenly distributed, being “in the room” was about access — who got invited, who had the clearance, who knew the right people. Now that information moves freely and AI can synthesize it for anyone, being in the room is about something else entirely: proximity to the kind of unplanned human interaction that generates ideas and connections you can’t get from your desk. The person you meet at the conference who changes your thinking. The hallway conversation that turns into a partnership. The weak tie that introduces you to someone three months later who changes your career.

Granovetter’s data says these are the connections that matter most for professional mobility. The serendipity research says these are the interactions that produce a third of all innovation. And the AI era makes both of those findings more true, not less, because the perspective advantage that weak ties carry has replaced the information advantage they used to carry.

Go to the thing. Talk to the person you don’t know. And when you do, don’t be shy about both listening deeply and sharing your distinct point-of-view.

Your Takeaway This Week

First, argue with an AI this week. Pick a professional opinion you hold strongly and spend fifteen minutes having Claude, ChatGPT, or Gemini argue against it. The point of the exercise is not to change your mind but to help you find the weak points in your own reasoning and refine your thinking. The goal is to walk into your next conversation with a point of view that’s been pressure-tested.

Second, have one real conversation with someone outside your immediate circle. This can happen at a meetup, at a coffee shop, at a kids’ activity while you’re both waiting. The venue is less important than the willingness to be genuinely interested in what someone else believes and genuinely willing to say what you believe back. Focus on having an engaging, deep, bi-directional conversation.

Third, notice what makes a conversation special. The next time you walk away from a conversation thinking about it hours later, ask yourself what happened. Chances are, someone said something that surprised you, and maybe it was something that came from their experience rather than from the general consensus. That’s the skill, and it has very little to do with AI and everything to do with being a person who has done the work to deeply explore their perspective.