The Investor's Dilemma
For decades, the best investors operated on a simple framework: find great teams, evaluate market size, assess the business model, and place a bet. Due diligence was a process of meetings, spreadsheets, reference calls, and pattern matching against past successes.
That framework is not wrong. But it is incomplete for the world we are entering. AI has introduced a new variable that most traditional investors are not equipped to evaluate: the technology itself. Not as a product feature, but as a structural advantage that can make or break an entire business thesis.
Why Operators Make Better AI Investors
The best AI investors I know are operators first. They have built products. They have deployed models. They understand the difference between a demo that impresses at a conference and a system that works reliably at scale. This distinction is everything.
A demo can be built in a weekend. A production system that handles edge cases, maintains accuracy over time, and integrates into existing workflows — that is the hard part. And if you have never built one, you cannot evaluate whether someone else's claim of having built one is real.
In the AI era, the most dangerous investor is the one who cannot tell the difference between a compelling demo and a defensible product.
What Changes in the AI Investment Thesis
Several things about investment evaluation change fundamentally when AI is at the center of a business.
Team size becomes a weaker signal. A two-person team with the right AI infrastructure can now outproduce a twenty-person team using traditional methods. The question is not how many engineers they have, but how effectively they use the tools available to them.
Moats look different. The traditional software moat — network effects, switching costs, data advantages — still matters. But AI introduces a new kind of moat: the speed at which a company can learn from its own data and improve its product. This creates a flywheel that is nearly impossible to replicate from the outside.
Revenue timelines compress. AI-first companies can reach meaningful revenue faster because they can build and iterate faster. This means the traditional milestone-based investment approach — seed for product, Series A for growth, Series B for scale — may need to be rethought. Some companies will be profitable before they would traditionally raise their first round.
The Data Question
Every AI pitch includes a slide about data. We have proprietary data. We have unique data. We have more data. Most of the time, this claim does not survive scrutiny.
What matters is not the volume of data but the quality and the exclusive access. Public data — no matter how much of it you have — is not a moat because everyone has it. Proprietary data that is continuously generated through your product's usage — that is a moat. It means your model gets better every time a customer uses it, and no competitor can replicate that improvement without building the same customer base first.
As an investor, the data question I always ask is not how much data do you have, but how does using your product generate data that makes your product better. If the answer is clear and defensible, the investment thesis is strong.
Valuation in a World of AI Deflation
Here is the tension that nobody in the investment world wants to talk about openly. AI is deflationary for software businesses. The same product that required a team of twenty and two years to build can now be built by a team of five in six months. This means the replacement cost of any software product is dropping rapidly.
This has profound implications for how we value companies. A high valuation based on the cost and time it took to build the product is no longer justified if someone can rebuild it in a fraction of the time. The value has to come from somewhere else — from the customer relationships, from the data flywheel, from the brand, from the operational expertise that cannot be replicated by code alone.
Smart investors are already adjusting their models. They are placing less weight on technology uniqueness and more weight on distribution, data moats, and the speed of the team's execution cycle.
Where I Am Investing My Attention
I look for three patterns when evaluating AI-era opportunities.
First, companies that use AI to enter markets that were previously uneconomical to serve. These are not sexy Silicon Valley markets. They are mid-market businesses, regional industries, and specialized professional services where the existing solutions are either too expensive or too generic. AI makes it possible to serve these markets profitably for the first time.
Second, companies where the founder has deep domain expertise in the problem they are solving. AI is a tool. Domain expertise determines whether you are using that tool to solve the right problem. The best AI companies I have seen are built by people who spent years frustrated by a problem before AI gave them the means to solve it.
Third, companies that are already generating revenue. In an era where building is cheap, the true test of a business is not whether you can build a product but whether anyone will pay for it. Revenue is the ultimate validation, and AI has made it possible to reach revenue faster than ever.
The Convergence
The line between operator and investor is blurring. The best investors are building. The best builders are investing. This convergence is not accidental — it is a natural consequence of a world where understanding the technology is prerequisite to evaluating it.
If you are an investor who has never deployed an AI model, you are evaluating businesses you do not fully understand. If you are a builder who has never thought about unit economics and defensibility, you are building businesses that may not survive. The future belongs to people who can do both — who understand the code and the capital, the technology and the market, the product and the spreadsheet.
That is where I operate. And I believe it is the only place worth being.