We've been talking to a lot of private market investors recently. I expected AI adoption on steroids from the same people who fund AI. From those conversations, we can split adoption into three distinct levels. Most firms can place themselves in one of these categories without much thought.
The three levels of VC AI adoption
The basic user
You use consumer-layer apps (ChatGPT, Claude, Gemini) for daily tasks like summarization, basic analysis, and asking questions. You use Perplexity or other research tools for market analysis. Most of the time, you copy-paste text from emails and CRMs, explaining repetitive tasks over and over. If you're here, you know you could do much more. Shifting focus to fixing that is never on the agenda.
The tech-savvy intermediate
You have a deeper understanding of the AI landscape. You're probably building your own Agent Skills, connecting your tools via MCP, and using them every day with your own flavor. Vibe coding your way to it. As an engineer myself, I respect people who get their hands dirty to solve a repetitive task. But in this phase, every employee lives in an isolated silo of context. Real insights are never shared, workflows are highly individual, and sharing them is usually awkward.
The power user
You custom-build your AI layer. You might even have a dedicated team for it. You want a proprietary company brain, knowledge base, and centralized workflows, usually run through an orchestrator like OpenClaw to tie it all together. If you're here, ask yourself: is the infrastructure you're building worth maintaining in-house? Does your fund actually run on AI, or are you just running an engineering shop?
The AI-adjacent trap
Here's the real deal: each level thinks the next one is winning the AI transformation. The basic user thinks that if they connect their tools to MCPs, they are done. The intermediate user thinks that if they add a context layer for their agents, they've won the race. And the power user thinks that just pouring more resources into their infrastructure will make everything right. None of them are wrong. At any level, AI is doing real work that saves real time. But they have it wrong about what it actually looks like to run a complete, unified system. What they have is AI-adjacent: close to the thing, shaped like the thing, but missing the part that makes it the thing.
The ceiling of the advanced DIY fund
Building in-house comes with a ceiling. You can't reach a deep, scalable solution when building software isn't your core business. Maintaining a battle-tested core that doesn't break, while constantly adapting to daily changes in AI methodology, is hard. The maintenance cost in an environment where AI changes every day is simply too high.
Basic apps are too shallow. Intermediate hacks are too siloed. Custom power-user setups are too expensive to maintain. So how does an investment firm actually use AI to win?
It takes a shift in perspective: from using AI tools to running on AI-native architecture. In our next piece, we break down what that actually means for a fund's daily operations.