Author: WIRED
Compiled by: Deep Tide TechFlow
Deep Tide Intro: Venture capitalists are the biggest believers in AI, collectively pouring over $200 billion into the AI sector last year. But an awkward question looms: will AI end up disrupting the VCs themselves? A platform called ADIN is already using AI Agents to replace human analysts for investment due diligence, completing in one hour work that used to take days or weeks. An even more existential threat is another layer—when AI makes the cost of starting a company plummet, founders might not need VC money at all. This article interviews several prominent VCs, presenting the real divisions and anxieties within the industry.
Full Text Below:
Last fall, as venture capitalists were flooding into the AI sector with record amounts of money, a group of investors gathered to evaluate a new project. The company was called Infinity Artificial Intelligence Institute, which made software to automatically tune AI models, making them faster and cheaper. The founding team looked solid, and the market was expanding rapidly. Half the investors were cautious, the other half smelled money. One called the deal an "absolute banger".
The company was real, and the $100,000 seed round these VCs invested in was real. But the VCs themselves were all AI Agents, part of a new platform called ADIN (Autonomous Deal Investing Network).
ADIN, launched in 2025, uses AI to replace human analysts in venture capital deals. Input a startup's pitch deck, and it outputs a detailed analysis of the business model and founding team, a list of due diligence questions and compliance risks, a TAM estimate, and a suggested valuation. ADIN has over a dozen different Agent investors, each with a unique persona and investment thesis. Tech Oracle looks at the underlying technology, Unit Master evaluates financial fundamentals, Monopoly Maker seeks market monopoly opportunities roughly based on Peter Thiel's style. When most Agents are bullish on a project, they recommend how much capital ADIN's fund should allocate to the deal. The whole process takes about an hour, whereas analysts at a VC firm typically need days to weeks.
"The venture capital game doesn't have a high success rate," said Aaron Wright, co-founder of Tribute Labs, ADIN's parent company. The current approach—a gut-feel, intuitive way of judging who will be tomorrow's great unicorn—has only about a 1% chance of hitting a "home run" (i.e., a project returning 10x+ the invested capital). Three-quarters of VC deals don't even return the principal.
In Wright's view, AI models can dramatically improve these odds. He believes venture capital is entering its own Moneyball era, where quantitative methods will surpass human intuition, and everyone will start hitting more home runs. "These systems will increasingly be able to weed out bad projects and focus on more successful ones, while also lowering the operating costs of these firms," Wright said. He believes that within a few years, AI Agents could become the world's best venture capitalists.
And then? "Sand Hill Road might not exist anymore."
No group is more bullish on AI than venture capitalists. They collectively invested over $200 billion into the AI sector last year. The advances in AI models have changed how investors view almost every company and every industry. Khosla Ventures founder Vinod Khosla recently predicted that AI will replace 80% of job duties by 2030. But many venture capitalists seem to underestimate the degree of impact AI will have on their own jobs.
Marc Andreessen—star VC and co-founder of Andreessen Horowitz—said on his podcast The Ben & Marc Show that after AI does everything else, venture capital might be "one of the last few areas where humans are still doing it." He sees the job as not just writing checks, but also choosing the right idea and the right people at the right time, and then guiding them to success.
"It's not a science, it's an art," Andreessen continued. "If it were a science, eventually someone would be able to dial it in and get it right eight times out of ten. But the real world doesn't work that way. You're in the business of contingent events. It has an ineffable quality, a factor of taste."
Many of the VCs I interviewed for this article held similar views. Keval Desai, managing partner at VC firm Shakti, compared early-stage investing to "picking Michael Jordan out of kindergarten." An early-stage project has no product, no revenue, only potential. "You can have all the compute, all the algorithms, but without data, there's nothing to analyze," Desai said. (Though he admitted he occasionally asks Gemini to "role-play as a VC analyst" for opinions on unfamiliar markets.)
Brian Nichols, co-founder of Angel Squad—an angel investing network associated with early-stage VC firm Hustle Fund—told me he wouldn't trust AI to do the "screening" work in investing. Ultimately, VC is a relationship business: it's about who you know and who you can vouch for personally. At the same time, he thought AI could probably replace other parts of the job. When we spoke, he had just returned from a Hustle Fund offsite where a partner had built a tool using Claude Code to triage founder emails. "We spend hours every day replying to founder pitches," he said. "That time could probably be spent elsewhere." Aydin Senkut, founder and managing partner of VC firm Felicis, told me he believes most VCs are experimenting with AI in some way to stay competitive. His firm is currently experimenting with using chatbots to write investment memos, improve deal sourcing, and help partners "score" founders.
Projects like ADIN attempt to automate more of the underlying work. The due diligence process—where investors investigate a project's feasibility, risks, and growth potential—is one of the most time-consuming parts of venture capital, especially when considering companies in emerging markets. ADIN compresses this step to minutes, quickly flagging regulatory or compliance issues that could scuttle a deal. When evaluating a mining technology company, ADIN flagged a series of export control regulations and cross-border data transfer issues. "These aren't questions most investors would think to ask," said Priyanka Desai, a partner at ADIN. She added that AI "doesn't get tired, doesn't have blind spots due to inertia, and can surface those long-tail risks that are easily overlooked."
For now, humans still do a few things. First, ADIN's deal flow comes from a network of venture capital scouts. Although ADIN has LPs funding it like a traditional VC fund, it offers scouts an unusual economic incentive—scouts can get 50% of the carried interest, which is usually profit reserved for GPs (General Partners). "It's basically giving GP-level economics to a person for just submitting deals and leveraging their network," Desai said.
Humans are also responsible for the "last mile," including meeting the founders and ultimately deciding whether to write the check. "We know these systems aren't perfect, so we need a double-check," Wright said. The AI Agents can sometimes be overly aggressive in their recommendations: he showed me one project that all the Agents loved, but ADIN decided not to invest after meeting the founder and discovering issues with existing competitors.
On the other hand, Wright said he has also used ADIN to evaluate some companies that have already raised over $20 million, some of which were unanimously disliked by ADIN's Agents. "The challenge for us is figuring out if this is accurate or a misjudgment?" he said. In some cases, investors might have fallen into a common human trap: hyping a project or founder based purely on gut feeling.
Whether AI systems can outperform investors is one thing. But there's another existential threat: the same AI technology that makes venture capital work faster and more efficient is also making it easier and cheaper to start a software company. Over the past decade, much of the money in the VC industry came from SaaS. But a project that once required a $2 million seed round to hire a specialized engineering team might now achieve the same product velocity with a few vibe coders and less than six figures in funding. The math of writing large checks no longer adds up.
Until recently, only a tiny percentage of unicorns were bootstrapped. According to SaaStr, which monitors SaaS companies, the average software unicorn raised $370 million. Now there are companies like the AI image generator Midjourney, which reached unicorn status with a core team of just a few dozen people. (According to the latest data from Pitchbook, Midjourney has about 100 employees. Court documents from a copyright lawsuit show the company has annual revenue exceeding $300 million. Midjourney did not respond to WIRED's request for comment.)
This scenario—where some founders simply don't need venture capital anymore—is the one most likely to strike fear into the hearts of venture capitalists. "This is the existential threat," said Angel Squad's Nichols. "The money is there, but the founders don't need it." Perhaps AI won't directly replace investors, but it might make those investments unnecessary.
Beyond robotics, biotech, or other hardware-type companies, there might soon be fewer startups needing the kind of large funding rounds on which the venture capital industry was built. This could return the industry to its origins: a small, specialized field bridging the gap between scientific breakthroughs and commercial application. (The giant companies building foundational models are still here, and they will likely continue to take VC money to pay for astronomical compute, data center, and employee costs.)
If starting up becomes cheap, we might see the industry shrink rapidly. This could put investors out of work in another way: not replaced, but their business model replaced. "If these funds are sitting around with nothing to do, scrambling for the very few deals that actually need funding, that creates another problem," Nichols said. "That's what keeps investors up at night."
