The question isn't whether to act on AI. It's whether you're willing to admit that what you're doing isn't working.

I've been building with AI every day for the past several months. Custom apps in Claude Code. Automations. Agentic workflows that handle real work while I'm heads-down on something else. I'll be honest: it's the most energized I've felt about my work since the early web3 days and we were building Holder. Everything was new. We were laying the infrastructure for a whole new industry in real time. Nobody knew what web3 was going to look like in five years, but we knew it was worth building. We were flying the proverbial plane while building it.

That feeling is worth naming. Because it's exactly the feeling that's missing for most of the innovation leaders we talk to right now. And the gap between people who are in it and people who are watching it is widening faster than most people realize.

You've seen the posts. Someone built an entire product in a weekend. A founder launched a company in two weeks. Anthropic adding billions of dollars in revenue faster than you can count. You read them, you feel something that isn't quite anxiety and isn't quite inspiration, and then you go back to your calendar.

That feeling is real. And it matters more than you might think.

The Structural Problem No One Wants to Say Out Loud

Here's what I keep noticing when I talk to others — especially those in large enterprises. It's not that they don't get what's happening. Most of them do. They're reading about agentic AI. They're watching demos. Some of them are even experimenting quietly on the side, exactly the way I was in the early days of web3.

The difference is what happens when they try to move. And that's where the structural reality sets in.

IT governance, legal review, risk management, procurement cycles. These systems exist because you're working in a multi-billion-dollar business, not a startup. They protect real things. They've saved you from real mistakes. The problem isn't that they exist. The problem is that they are now the single biggest obstacle between you and whatever window is open right now.

We've talked to corporate innovation leaders across dozens of the world's largest organizations, and the pattern is almost identical. Eighteen months into trying to run meaningful AI experiments. Approvals pending. Pilots stalled. A mountain of management consultant decks. A team of talented people who are frustrated and starting to wonder if anyone actually wants them to succeed.

But here's what most people aren't talking about yet: the risk isn't just moving too slowly. New research out of MIT Sloan is starting to document that organizations deploying AI the wrong way, passively, through existing workflows, with workers delegating rather than directing, are actually making their people worse over time. Not in some theoretical sense. In measurable ways. You add the tool, productivity bumps in the short run, and eighteen months later the team can't do the underlying work as well as they could before. The researchers call it the "augmentation trap" (the tendency for AI to boost short-term output while quietly eroding the expertise teams need to do the underlying work well).

So you have two risks now: missing the window, and catching it wrong.

It's Time to Change the Conversation

One entrepreneur. 1,100 AI agents. Nine companies. Two months.

That's not a hypothetical. That's what's happening right now inside an agentic venture studio. Ben Lewis, my colleague running our OneHealth Studio, built and runs a swarm of more than a thousand agents organized into a chain of command, each narrowly specialized, working around the clock. He wakes up every morning and reviews what they produced overnight.

This is not just "one person doing the work of hundreds." One person (an entrepreneur) with the right incentives, the right autonomy, and an army of agents can compress timelines that used to take years into months. The human judgment is still the irreplaceable part. The agents amplify what that person can do. The go/no-go decisions, the strategic direction, the quality filter: that's still human. It has to be.

Your innovation team on the other hand, has thirty people and hasn't shipped in eighteen months.

Ben wrote about how this works in detail. If you want to understand the actual mechanics, start there. What I want to focus on is what it means for how you're thinking about your corporate AI strategy.

What Building in This Moment Actually Feels Like

I keep coming back to web3. Specifically, to a piece Matt Huang at Paradigm wrote in 2023 called The Casino on Mars. His frame was that crypto was like settling a new planet. Early settlers were a mixed bunch: explorers, speculators, researchers, builders. The infrastructure was rough. A lot of what got built was noise. But underneath the casino, real infrastructure was going up.

What he captured perfectly was the paradox of building in a moment like that. The speculation and the chaos aren't separate from the progress: they're how the progress gets funded and built. The casino bootstraps the infrastructure. You can't always tell in real time which is which.

That's where I am with agentic AI right now. I'm building custom tools in Claude Code for work I used to do manually. I'm running automations where agents handle the execution and I'm the human in the loop reviewing, approving, redirecting. I'm experimenting with architectures that I don't fully understand yet, and I'm learning faster than I have in years because the feedback loops are so short.

Some of what I'm building right now will matter. Some of it won't. But I'm learning things I couldn't learn any other way, exactly like the early days of web3, when building was the entire point, even before we knew what we were building toward.

The people who aren't in it are reading about it. There's a meaningful difference between those two things, and the gap compounds over time.

Your Corporate AI Strategy Needs an Overhaul

Here's what I keep hearing, underneath the conversations about AI strategy and roadmaps and investment theses: "We know we need to do something. We just don't know how to do it inside the organization."

So let me ask the harder version of that: What if inside the organization isn't the right place to do it?

That might sound like giving up. It isn't. Corporations have always spun ideas out into separate entities to give them room to breathe. That logic didn't disappear because we're talking about AI. It got more urgent. The difference now is that outside doesn't mean slower. In some cases it means orders of magnitude faster.

The question isn't whether to have a corporate AI strategy. It's whether the answer to that strategy can actually be built within the constraints you're operating under. For some things, it can. For others, the honest answer is that your internal constraints will kill it before it can prove anything, and two years from now you'll have a team that's simultaneously slower and less capable than when you started. That's the augmentation trap in practice, playing out at the organizational level.

Acting Doesn't Require Understanding Everything

Acting doesn't require understanding every detail of agentic AI, agent swarms, multi-agent architectures, or orchestration infrastructure. It requires making a decision about where to place the bet.

Some of that work can happen internally, and it should. But for the ideas on your shelf that have been waiting for bandwidth, for the opportunities you know are real but can't get traction on internally, there's a different path. One that routes the work outside the organization entirely, to a venture studio operating at startup speed with the infrastructure already in place.

The window here is real. I can't tell you exactly how long it stays open. What I can tell you is that the companies who move in the next six months will have proof points to show for it. The companies still in governance review will be watching them present at the conference.

You already know the answer. You're just waiting for permission to say it.

If you're ready to have that conversation, reach out. Or start by reading what venture building at this pace actually looks like in practice.