Automating Our Most Human Industry: Venture Capital

TL;DR

Automating venture capital (VC) poses unique challenges due to its reliance on private markets and human-driven decisions. While innovative platforms and crypto initiatives demonstrate potential, true automation requires:

• Carefully designed investable assets
• Multi-asset models
• Ecosystemic strategies that integrate different asset classes

I write this not to plant my flag in the ground, but to ask the question — what’s next in automatic, quantitative, data-driven, AI venture capital? What is this budding ecosystem becoming? Ecosystemic thinking makes the task of understanding the whole space practically impossible, but eternally valuable to pursue. So can I ask those reading to tell me more about any initiatives they know about solving one element of the puzzle, any pieces I’ve missed, and also their own pursuits in filling this in? This is one of my great passions, so I’m always keen to learn more.

Recently I got into a lovely conversation about the possibility of successfully automating Venture Capital with my cohort at the Venture Institute. It’s a topic I’ve been pounding my head over for some years now, and through tough conversations, I think I have some thoughts on the matter.

We all see automation in finance — most people at least know of trading algorithms by now. But the issue is that almost all the automation we see is in public markets, not private markets, where VC squarely sits.

Public markets are well automated because they have been formalised to create an even playing field for information — it is literally a crime to conduct insider trading over the stock market. This means that the name of the game there is what you do with the information, not what you can gain.

The opposite is true for private markets, where information asymmetry rules the day — the game here is to gather as much information as fast as possible to make investments or build businesses and outcompete your competitors. At the end of the day, private markets play at the most basic unit of innovation — the humans doing it.

VC is as deep into private markets as you can get. I mention this because the difference in markets reflects the two sides to automation — building the model to process your data, and automating the data collection — and you need that data in place to process anything.

The problem with this is that getting data that is useful for any models is an extremely long and laborious process, and VC relies not on having the data, but making the right decisions — human beings are extremely well-built to make correct snap decisions without full comprehension of why. Now do you see the problem?

The moment you put the barriers of any bureaucracy that automation demands, you slow down the information capture process, so without being smart about it, you completely undercut your own competitive advantage, and you lose the chance to automate anything. So you have to be smart.

There are platforms for VC — you’ll have heard of some, not heard others. Here’s a couple:

• Crunchbase — offering company and investor profiles (recently developed AI-driven predictive analytics forecasting startup funding events with 95% accuracy)
• PitchBook — with industry trends and deal data (reporting AI-focused companies attracting over 50% of global VC funding in late 2024)
• Decile Group — synchronising venture operations
• shipshape.vc — searching broad investor databases
• Affinity.co — automating CRM systems by capturing and organising interactions

In performing these functions, they propose that the time taken to sign up and use them is less than the time taken to do things without them, and so both profit and information collection is automated via market functions.

This however is changing the nature of VC — with every standardised platform, more information becomes an even playing field, and the earlier the “public” the market gets. Is it really the same asset category when this happens?

What separates private equity from public equity?

Consider cryptocurrency and blockchain innovations. These are, ostensibly, private market investments; however, the presence of crypto exchanges makes many have to adhere to certain reporting and transparency standards. This forms an interesting bridge between the two, and automation in this space is well expressed both with:

• Quantitative algorithms — some of the most interesting and advanced in the world, including those used by Numerai and Polychain Capital, and
• Decentralised Autonomous Organisations (DAOs) — such as VitaDAO.

A DAO is effectively a fund that leverages community intelligence by rewarding investment functions with stake in an investment via paying the cryptocurrency associated with that investment. The autonomy lies in the autonomous stake assignment for completed tasks and the function of the community — the coin has rules in its operation that are designed to self-sustain independent activity (often governed by the community). There’s a catch, though — crypto is a tremendously corrupt sector.

Coins can well benefit from pyramid schemes — coin purchase is coin use; even Ponzi schemes can be very profitable if the schemer can get away with it, and they do, leading to perverse incentives. In proposing a publicly transparent market, without being truly public, savvy players can selectively withhold information without clarifying it is withheld, leading to fraud. Even outside of fraud, insider knowledge about a deal is still the most important factor in making sound investments. It hasn’t truly solved the automation problem.

So many people are doing so many incredible things trying to solve it, for example:

• Backtesting AI models for VC and report high accuracy (such as Revolut or Earlybird Venture Capital), or
• Creating platforms that automatically identify startups (like Crowdcube), or
• Creating algorithmic decision-making systems, enabling rapid evaluations (like Correlation Ventures).

But I also know that they don’t rule the world yet. The big challenge of VC is that it’s not stationary — by definition, it breaks the status quo. The trick then is correctly processing and incorporating data on the many steps of business operations, structuring it, presenting it in such a way that a human being can quickly act, and wrapping this together with incentives to participate.

How can we connect disparate assets to complement one another?

The potential for scalable automation lies then in business model innovation, with strategies that incorporate multiple asset classes, or in securing information from common platforms across many horizontals. To have genuine potential for automation, you should:

• Promptly process data in some platform
• Integrate this with OSINT to figure out how to integrate it into another platform
• Then partner with that other platform to share models or data (such as Carta’s cap-table management automation or AngelList’s investment syndication platform).

This is where the asset class problem comes in though — good luck getting quick, accurate, current data on VC; it, by definition, doesn’t exist. What thinking must be done to define this novel asset class?

Consider this: the goal should not be to automate VC, but to automate the use of the assets one has built up — knowledge, connections, organisation, dealflow, and so forth. This act is, in truth, figuring out how to apply the techniques of VC to an automatable asset class worth your time. I’ll give you my suggestions:

• Design investable assets resistant to corrupt practices
• Identify common elements between VC and other assets which can mutually inform and observe each other
• Develop multi-asset models that feed into one another — low involvement with fast returns building to high involvement long returns are the target

What happens when we truly combine such different perspectives? Can we do it?

Essentially, the next step is to build the links between the organs of investment. If we each do one, we may eventually stumble upon a full organism, the cousin of Homo Economicus, known as Homo Financius. We are moving up a stage on the evolutionary tree, and it is very exciting. Please tell me more about your thoughts, your initiatives, and how wrong I am in the comments — these discussions are why I wrote this in the first place.