About 9 months ago I made the shift from being a full-time VC to being a full-time Founder, and since then I’ve personally managed a few fundraises for formation stage Startups. It was a strange feeling being on the other side of the table and I couldn’t help but analyze the work that was being done by the people in my former role. I’ve shared my observations in former posts, but over the course of the past few months my mind has been drawn to the meta question about whether the “work to be done” in the VC industry is going to be disrupted in the same way that AI startups are disrupting other knowledge-based professions. The irony makes me smile.
The VC industry is supposed to be the catalyst for disruption and innovation. The VCs have been excited to displace lawyers with AI contract tools and accountants with agentic driven workflows. And now the industry might be bankrolling the very agents that could sideline their colleagues.
Answering this question is really a thought exercise based on an understanding of the “work to be done” and where the edge comes In VC. The ultimate goal is to source and spot Founders who can describe futures that rewrite the rules of an industry and then capture massive amounts of value (profit) if they can make their vision actually come true. But not all the steps to get there are immune to AI. If fact, some of them may be enhanced by it.
As part of this thought exercise, I wondered where and when the disruption would start and came to the conclusion that the more mechanical, “math-based analysis” parts of the job that come with evaluating later stage companies is obvious. But the more interesting part of the thought exercise led me to believe that an AI-driven seed firm might be able to handle chunks of the VC job better than the ecosystem does today. It could be faster, fairer, and operate with less baggage. I’m not convinced of my own argument, but the “what if” exercise is worth sharing.
What comes next is a breakdown of the core concept followed by speculation about how this could plays out over the next few years. Call it speculative fiction grounded in today’s trends. But I am convinced some experiments will pop up soon, and if they pan out then it’s not a matter of “if”, it’s “when”.
Start with the human flaws we all know exist.
Humans are built to think in terms of pattern-matching because it’s an amazing survival skill. Evolution has given us a gift for identifying and dodging threats that’s seared into who we are as a species. Fire hot, don’t touch is a basic form of pattern matching that’s served us well. But pattern-matching can be a curse in identifying breakthroughs because definitionally they don’t match the patterns of the past. Most humans struggle to unlearn and our knowledge base is a product of our own experiences. A scar from a failed investment or living through “crypto winter” has an impact on how a VC “hears” every pitch. We chase familiar patterns that echo our past successes and we run away from familiar patterns that elicit PTSD.
Yet the mega-returns flow from Founders wanting to manifest worlds that don’t exist yet. Evaluating their visions as “investment opportunities” requires dissecting industries, handicapping odds that live in the unknown, and judging the leadership and judgment skills of the Founders. The best VCs are skilled at doing these things well. But even the best VCs are shaped by their histories and often miss outliers that shatter old paradigms. We routinely pass on outliers because they don’t fit the patterns we’ve internalized.
Can a seed stage AI VC firm do better?
Imagine a world where a VC firm trains a system with all the deal histories from the past few decades along with outcomes and battle-tested insights. This will become a foundational knowledge base that AI can use to discover the patterns of the past and use as part of a multi-faceted evaluation process. Add to this a 150 IQ AI agent that specializes in drawing out the unique insights from a Founder’s vision and evaluating them. Generational companies are built based on a description of a future state that if it comes true allows the Founder to build and scale an amazing business. The 150 IQ AI agent can be taught how to think in terms of future states and evaluate the probabilities of various drivers and key events happening.
Once these agents are trained, they can be put out into the world to run the show. Besides the core evaluation agents, picture other agents that comb the endless deal flow from AngelList, YC, GitHub, LinkedIn and even relationships it builds with Principals and Partners at other firms. It would be 100% open for business to any and all applicants, so the inbound volume could also be massive.
An interaction agent could handle live pitches via video, drilling the Founder with tailored questions based on the analysis produced by the analysis agents, all the time refining its questions and flagging risks. Decisions can be made quickly, especially if the terms are fixed and simple. Imagine a tuck-in check for a $500K SAFE with a 24-hour expiration window to avoid being the “funding you choose when you can’t get the funding you want”. A close would be triggered based on a qualified round coming together.
Tuck-ins are a good place to start because they can be kept simple. Low stakes and high throughput is perfectly suited for the characteristics of the seed stage ecosystem. Speed and certainty could be an attractive value proposition to Founders. No ghosting and no drawn-out dances. Just capital to build.
And the AI would share its investment memo with the Founder. Pitches could be tweaked and it could be made available by the Founder to help get other Investors over their concerns, turning your firm into a source of “free diligence”.
This system doesn’t describe a total replacement for humans in the process, especially if the VCs want to spend time with the Founders to get a personal read. There are special Founders with raw talent and charisma that could be missed by the AI agent swarm. But they could collapse a long, drawn-out process into a single pitch to the swarm plus a single meeting with a VC Partner who’s digested the output of the AI and is prepared to make a call. A hybrid system probably wins (for now).
There’s another trend that makes this future not just an interesting thought experiment, but possibly a necessary evolution. Because AI has basically created an infinite number of software engineers, we’re poised for a Cambrian explosion of new Startups being formed. Not 2X more than in the past. Not 10X more. Orders of magnitude more. Most of the Statups will be sheer and utter nonsense but it still requires time and effort to sift through all the pitches. This means that a Seed stage VC’s job might end up being to find “the cream of the crap” which will be time consuming, frustrating and chock full of Type II errors. Efficiency gains at the seed stage might become a necessity.
Now for some fun speculation. Let’s timeline how this could play out.
2026-2027: The Tuck-In Experiments Launch
With near certainty someone’s going to do this soon. Maybe it’s a scrappy solo-GP who became a centi-millionaire from their time at OpenAI or an a16z offshoot that tests the waters with a $20-$30MM pool of capital. They focus on tuck-ins first, making dozens of small bets to prove that the model works. Founders flock to it out of curiosity, because of the names involved and for the no-BS process. Early wins quickly show that the agents can spot Startups that downstream capital likes.
By late 2026 the major AI Foundation companies take note of the successful experiment and lean in. They’re comfortable with agents and recognize that fueling the Startup ecosystem is a meta-investment in their own tools.
Early wins include the Startups that typically would have been funded but now start to include underrepresented Founders from non-traditional backgrounds and geographies. Skeptics scoff at “robot VCs,” but the data suggests it’s working. By year-end, a couple of new AI funds launch and attract more LP capital.
2027-2028: Efficiency Purge. Analysts and Principals Get Automated.
With tuck-in proof in hand, the AI firms start to scale. But elsewhere in the VC ecosystem, agents start to replace junior roles lock stock and barrel. Analysts screening inbound deals? That work has been handed off to an agent. Principals running the diligence process? The work is better orchestrated by an agent with tasks handed out to specific agents and human decision makers. Deals start to close in days, not months, while the expense base of the VCs plummet. Fewer people and less travel means that management fees go a lot farther and could be reduced to allow LP funds to be “invested” instead of “spent”. The result is a more efficient organization that allows for more shots on goal.
Most firms start to contemplate if they should become a “hybrid firm”. Partners focus on relationships and the complex work of guiding Founders and agents handle the quantitative and volume pieces of the business.
Bubbles form in hot theses, but agents adjust faster than humans and can spot the bubbles before they burst. The lack of FOMO and their ability to unlearn biases overnight cools overly hot sectors quickly. Job shakeups in the VC industry hit hard and Principals leave to work for portfolio companies or pivot to “AI trainer” gigs. By late 2027, the top funds lean heavily into agent integrations and pressure the laggards to keep up.
2028 and Beyond: Transformation
Time only improves the performance and validates the thesis. The foundational intelligence underneath the system improves and is better at analysis and prognosticating than all but the best of the best VCs in the industry. Superintelligent agents tackle uncertainties with simulations and can understand the implications of major shifts like quantum computing and autonomous robots. LPs decide to buy in to the concept of hybrid VC firms big time. The AI Foundation companies become large LPs, then sovereigns and endowments follow based on the results of the pilots.
Humans handle boards and politics. Agents focus on deal flow, analysis, draft terms, cut checks and arm the humans with up to date synthesized information as the investments mature. Full disruption won’t take place because the intangibles are real and valuable. But more outliers get funded and diverse unicorns get created. Incumbents adapt or struggle to raise LP capital. New AI-native firms start to dominate the early-stage ecosystem.
Will it happen?
Here ends the thought experiment and hopefully it was an interesting one. Nobody knows if the future described above will come true, especially in such a short timeframe. But what is highly likely is that pieces of it will come true, and probably faster than the VC industry would like. This isn’t AI creating an extinction event for the VC ecosystem. It’s just basic evolution at work. VCs are funding forces that will impact them in similar ways to how they’re impacting every other knowledge-based industry. There are no exemptions.
I'm convinced that an AI would do better than most VCs but I am also convinced that a monkey would do better than most CEOs.
But in venture we don't care about "most", we care about the outliers than return the fund.