I invest in AI because I build AI. That order matters, and it shapes a thesis I am willing to put a fund behind.
Here it is in one line: in the agentic era, value does not accrue to whoever has the best model. It accrues to whoever owns the workflow, the data, and the distribution.
Let me defend each.
The model is a depreciating asset.
Foundation models are extraordinary and they are also a commodity in the economic sense: capital-intensive to build, cheap to rent, and improving on a curve you do not control. If your company's advantage is that you have wired up a slightly better model than the competitor, that advantage has a shelf life measured in months. I have watched too many "AI companies" whose entire moat was an API key. When the underlying model got better, or cheaper for everyone, the moat evaporated. I underwrite against that.
Workflow ownership is durable.
The companies that last own a workflow deeply enough that the model is just one component. They understand the messy, domain-specific process (underwriting a policy, closing the books, staffing a shift) well enough to redesign it around agents, with the human gates in the right places. That understanding is hard-won and hard to copy. A big-model incumbent can match your API. It cannot easily match ten years of knowing exactly how a niche process breaks.
Proprietary data compounds.
Everyone has the same models. Not everyone has the same data. The teams that win capture data their product uniquely generates and feed it back into the loop, so the product gets better in a way a competitor starting today cannot shortcut. This is the flywheel I look for: does using the product create an asset only this company accumulates?
Distribution decides who even gets to compete.
The best AI product with no path to customers loses to a mediocre one that is already in the room. This is also, candidly, our unfair advantage as investors. We source from a network we co-built over eight years: 200,000+ members, thousands of founders, and access to Fortune 500 technology leaders. That network has already surfaced founders who went on to raise, collectively, hundreds of millions. We see AI-native founders before the market does, and we can put the ones we back in front of their first enterprise customers. Distribution is not a footnote to the thesis. For an emerging fund, it is the whole edge.
So what do we actually back?
AI-native software in categories where a founder who owns the workflow can outrun a big-model incumbent: future of work, fintech, digital wellbeing. Founders who are building AI-native, not AI-flavored: agents designed into the product, not a chatbot bolted onto an old one. And founders where our network is genuinely the accelerant, not a nice-to-have.
I bring one more thing to the diligence that most investors cannot: I ship this stuff. When a founder walks me through their architecture, I can tell in about ten minutes whether it is a real AI-native system or a repackaged one, because I have built both. That is not a better spreadsheet. It is a different kind of underwriting: an operator's eye paired with institutional discipline I learned running technology for a platform that supported hundreds of billions in investment.
The best model will not win. The best-owned workflow, fed by proprietary data, delivered through real distribution, run by a founder who understands all three: that wins. That is what I build toward, and it is what I back.
If that is the kind of edge you want exposure to as an investor, my door is open.