Charles Yang, Founder and CEO of Vibe – Interview Series

Charles Yang, Founder and CEO of Vibe, is an entrepreneur focused on rethinking how teams capture, retain, and build upon organizational knowledge. After co-founding a mobile gaming company that scaled to more than 800 employees and 50 million monthly players before becoming part of Tencent, Yang became increasingly interested in the challenge of preserving institutional memory as organizations grow. He later explored immersive collaboration technologies through Inlight Interactive before founding Vibe in 2017. Under his leadership, Vibe evolved from a smart collaboration hardware company adopted by more than 50,000 organizations into a broader “Contextual AI Workspace” platform designed to transform meetings, conversations, and decisions into structured, living organizational memory powered by AI.

Vibe is building what it describes as a Contextual AI Workspace, combining AI software with purpose-built collaboration hardware to help organizations preserve and operationalize institutional knowledge. The company initially gained traction through interactive smart boards used by enterprises and educational institutions, but has since expanded into AI-powered meeting intelligence, contextual memory systems, and physical AI devices like the Vibe Bot. Its platform is designed to capture conversations across meetings, documents, and collaborative sessions, then organize them into searchable, persistent workflows that teams can continuously build upon instead of repeatedly recreating context. The company’s broader vision centers on “memory-native” organizations, where AI acts less like a chatbot and more like an evolving organizational intelligence layer embedded directly into everyday collaboration tools.

You founded multiple companies before launching Vibe, including a gaming company that scaled to more than 50 million monthly players and was later acquired by Tencent. How did those experiences shape your belief that “organizational memory” would become one of the biggest unsolved problems in enterprise AI?

Running multiple companies, especially the gaming one, the thing that surprised me most wasn’t the technical scaling. It was the human scaling. You add a hundred people in six months, and suddenly half the company is making decisions based on conversations the other half wasn’t in. Meetings that recap last week’s meeting. Repetitive work because someone misremembered an action item. Monday mornings that start with “wait, what did we decide last week?”

I assumed this was a startup problem we’d outgrow. It isn’t. The bigger the company got, the worse it got, because context was being created faster than any system could hold onto it. By the time we sold to Tencent, I was convinced the real bottleneck in most organizations wasn’t talent or strategy — it was memory.

That’s what I wanted Vibe to fix. Not because organizational memory is some abstract concept, but because I’ve watched smart teams lose months of progress to it.

Apple proved that connected devices become more valuable when they operate as part of a broader ecosystem rather than as isolated hardware. Why did you believe enterprise AI would eventually need a similar ecosystem approach instead of relying purely on SaaS applications?

SaaS was never really built for the way work actually happens. Every tool owns its own slice — the CRM owns deals, the doc tool owns documents, the meeting tool owns meetings — and the employee ends up doing the integration work in their head, copying between apps, reminding each tool what the others already know. That works when work is task-based and lives inside a screen. It falls apart the moment a project moves between a job site, a hallway, a client call, and three different apps in an afternoon, which is most projects.

I came at this from years of building consumer products before Vibe, and Apple’s lesson stuck with me. They didn’t win because their individual devices were better. They won because the devices shared context. You moved between them, and the work moved with you.

Enterprise AI has to look more like that and less like another app you have to teach. Devices that capture context wherever work happens, all feeding the same place. The employee stops being the connector.

You describe Vibe as building a “Contextual AI Workspace” where conversations become permanent team intelligence. What was the key realization that convinced you this needed to be solved with physical devices like Vibe Board, Vibe Bot, and Vibe Dot rather than software alone?

The realization was actually pretty mundane. I’d watch people work and notice how much of the real thinking happened away from their laptops. Someone sketches an idea on a whiteboard. A site supervisor figures out a fix standing in front of the problem. Two execs make a real decision on the walk back from lunch. None of that is happening on a screen, which means AI software has no idea any of it occurred.

That’s the limit of software-only AI. If the most important moments of your workday aren’t on a screen, software alone can’t give you a real picture of the organization. You have to put capture into the environment where work is actually happening — the room, the site, the conversation. That’s what Vibe Board, Bot, and Dot are.

Physicality also matters because I wanted to build something beautiful. If we want this in the workspace every day, it has to be something people actually want there.

A lot of institutional knowledge still disappears in hallway conversations, whiteboard sessions, and informal discussions. Why do you think traditional enterprise software has struggled to capture this off-screen context effectively?

There are too many steps. Think about it: you’re in the middle of a conversation, something interesting comes up, and to capture it you have to pull out your phone, unlock it with Face ID, find your notes app, find the record button, hit it. That’s thirty seconds to a minute. By the time you’re recording, the moment is gone. The idea is gone. The other person has moved on.

The deeper reason enterprise software has struggled with this is structural. Software assumes the user comes to it. You open the app. You log in. You navigate to the right place. That model works when someone is sitting still at a desk. It doesn’t work in a hallway, on a job site, or in the five seconds after a meeting ends when the real conversation happens. The thing that needs to be captured isn’t going to wait for you to launch an app. That’s a hardware problem, not a software one.

That’s why we designed Vibe Dot the way we did. Press and record. The friction is gone, which means the capture actually happens.

Many AI companies are racing toward autonomous agents, but your approach seems centered around persistent memory and context first. Do you believe memory infrastructure becomes the missing prerequisite before AI agents can actually function effectively inside enterprises?

Yes. An AI agent without memory is like hiring someone who forgets everything between Monday and Tuesday. You wouldn’t trust that person with anything important, and you shouldn’t trust an agent built the same way.

Agents aren’t failing because the underlying models aren’t smart. They’re failing because they walk into every task with no context. They don’t know what was decided last week, what the customer’s history is, what’s already been tried and didn’t work. So they make confident, plausible decisions that are wrong in ways only someone with memory would catch. The same agent, given the same task on two different days, can make two different decisions, because there’s no continuity holding it together.

You can’t build trustworthy autonomy on top of nothing. You build it on top of memory. Get that layer right and agents become useful. Skip it and you’re just generating confident guesses at scale.

You’ve spoken about “memory-native organizations.” How different do you think companies will operate once AI systems can retain and connect years of organizational context instead of relying on fragmented documents and human recall?

The biggest change isn’t speed or accuracy in the abstract. It’s that a memory-native organization stops repeating itself.

Take onboarding. Today, when you hire or promote someone, it takes weeks to get them up to speed — finding documents, hunting down workflows, asking five people for context that was never written down. Most of what a new employee actually needs isn’t in any system. It’s in someone’s head. A memory-native company hands them everything that’s ever been decided, debated, or tried, the moment they start.

But onboarding is just the easy example. You can actually answer “why did we do it this way” instead of guessing. Handoffs between teams stop dropping things — sales to CS, design to engineering. When someone leaves, their context doesn’t walk out the door with them. Most companies relearn the same lessons every two or three years because the people who learned them are gone. That stops being a tax you have to pay.

Hybrid work accelerated demand for collaboration tools, but many companies still struggle with alignment and continuity. Do you think the next phase of enterprise AI is less about generating content and more about preserving collective organizational intelligence?

Yes. And I think most enterprises already know this, even if they haven’t named it yet. The content generation problem is largely solved. What’s broken is continuity. A decision gets made in a hallway, a client call reveals something important, a site walk changes the direction of a project — and none of it makes it into any system. The organization’s collective intelligence lives in conversations that disappear. The next phase is about capturing that layer, not producing more output from it. You can’t build real organizational memory on top of typed reconstructions of what actually happened.

There are obvious privacy and governance questions whenever workplace conversations are continuously captured and structured into institutional memory. How do you think enterprises should balance the benefits of persistent AI memory with employee trust and transparency?

This question has a simpler answer than most people believe. You need to be explicit about what’s being captured, who can see it, and what it’s used for. The discomfort people feel around this usually isn’t about capture itself, it’s about capture that happens without consent or clarity. Most people don’t object to a meeting being recorded — they object to not knowing it’s happening or where the summary ends up. What breaks trust is ambiguity. Enterprises that treat this as a policy problem rather than a culture problem will get it wrong.

Your company evolved from smart collaboration boards into wearable AI devices and ambient workplace intelligence. How do you decide when a workflow problem should be solved through hardware versus software?

Quick reframe: I think of these less as wearable AI devices and more as ambient hardware built for the environments where work happens — and that distinction is exactly how we decide hardware versus software.

Software can only work with what it’s given. If the context doesn’t make it in, software has nothing to act on. So when I look at any workflow problem, the question I ask is simple: where does the actual work happen, and can software reach it?

Most enterprise software was designed assuming the answer is yes — that the user will eventually sit down at a laptop and type the context in. For a lot of work, that’s not realistic. A contractor isn’t bringing a laptop to a site walk. A teacher doesn’t pull up an app between classes. A partner at a law firm isn’t typing notes during a client meeting. That’s where hardware becomes necessary. Software can’t reach those places. Hardware can.

Looking ahead five years, do you think enterprises will still primarily interact with AI through screens and apps, or will ambient AI devices and contextual environments become the dominant interface layer for work?

Screens will likely stay, but they’ll stop being the primary or only input layer. Right now, we treat the screen as both where work gets done and where we tell AI about the work. Those two things are going to separate. The screen stays valuable for output, for review, for decisions. But the capture side, the part where AI learns what you’re actually working on, that will move into your environment with the rise of ambient devices, wearables, and in-room hardware. Five years from now, the idea that you had to type your context into an AI tool before it could help you will feel like it belongs in a different era.

Thank you for the great interview, readers who wish to learn more should visit Vibe.