VCs rely on AI to find startups and decide on investments. André Rettherath, partner at Earlybird, knows the world of “data-driven VCs” like no other and gives founders essential tips.
André Retterath is a partner at Earlybird in Munich and a pioneer in the field of AI in the VC business. Getty Images / PM Images + that for his photo, collage: founder scene
It’s no longer a dream of the future: AIs decide whether startups get financing or not. Never alone and not in every case, but more and more VCs are using AI to make better investments, as they wish. In other words, those that bring in as much return as possible.
This presents founders with the question: How do I pitch to an AI? What do you have to do to be found by crawlers and considered good by algorithms? Which data is relevant for the machine? How do you impress the AI investor? And: Is that even possible?
André Retterath is a partner at Earlybird in Munich and a pioneer in the field of AI in the VC business. He has been pushing the topic of “data-driven VC” for years and has now published an annual report on global ones for the second time „Data-driven VC Landscape 2024“ published.
We’ll summarize what’s in there for you – and provide answers to the question of how startups should prepare for “data-driven VCs” (DDVCs for short).
What exactly is a DDVC?
Retterath has around 190 such DDVCs worldwide. He defines them as follows: In addition to analysts, investors and portfolio managers, there must be at least one programmer in the team and the company must demonstrably develop internal tech tools that improve the investment process. He includes, among others, Andreessen Horowitz, Atomico, Sequoia, Lightspeed, EQT and of course Earlybird.
Read too
These investors let an AI search for new unicorns – these are their surprising findings
35 percent of these DDVCs say that half of their deals are already flown by AI. “These funds believe that AI will enable them to capture data more comprehensively, process it more efficiently and ultimately make decisions more effectively. So that fewer people can do more work with better quality,” the report says.
There is potential for improvement at pretty much all points in the process: in sourcing, i.e. the search for startups, in screening, i.e. in the review of pitch decks and initial contacts, in due diligence, the closing of the deal and also later in the “portfolio Value Creation”, i.e. portfolio management, and exit.
Where is AI used in the investment process?
Currently, AI is most commonly used at the very beginning of the process, in sourcing and screening. Because that is where the strengths of artificial intelligence lie: it can sift through large amounts of data better, i.e. faster and, in the best case, more accurately, than a human. And: The AI has no bias, is not prejudiced. A problem that people, because of their humanity, can hardly get rid of: There are founders who are likeable to an investor – and others are not. Often certain similarities that both have play a role unintentionally and unconsciously. Be it the same gender, similar education, comparable traits.
According to the report, a third of VCs surveyed source more than half of the startups they later invest in using AI. She scours the internet for new companies and opportunities. Around 75 percent use AI in screening and sourcing. This means that the investment process – for these companies – is changing from “inbound” to “outbound sourcing,” according to the report. It’s not the founders who are looking for investors; rather, they have to let investors find them.
This raises three crucial questions for the editor of the study, André Retterath:
1. Where should startups be listed in order to be discovered online by a VC AI?
“The most advanced data-driven funds now pull data from everywhere it is legally possible,” explains the investor. The crawlers are constantly running through the network. They do this with two specific goals: on the one hand, to identify new companies, and on the other hand, to collect as much data as possible from companies that have already been identified.
“The spectrum for identification includes almost everything where companies leave digital footprints at an early stage – such as online forums for finding co-founders, Linkedin with information such as ‘Starting something new’ or ‘Stealth Mode’.” But also the commercial register for company registration and product websites like ProductHunt would be checked regularly. Pitchbook, Dealroom, Crunchbase, CBInside, Ventureexpert, Prequin are also mentioned as key sources in his report.
2. Which metrics, factors and data are crucial for AI?
“We differentiate between rule-based and statistical screening approaches,” explains Retterath. In the former, the rules-based approach, investors themselves determined what they wanted to see. For example, a focus on certain industries and geographies or employee growth and website traffic. “With the statistical approach, we train classification models to discover historical patterns of successful companies,” the investor continued.
“For example, in a quick-commerce company, founders with “execution-heavy” backgrounds such as consulting or investment banking and with training at a business school are more successful.” In a nuclear fusion company, on the other hand, founders with a High academic qualifications from a top university and papers that are cited as frequently as possible will be ranked better. “These are what we call moderator variables. For example, depending on the industry, the factors that make a company’s success more likely change.”
3. What bothers the AI and reduces the chances of an investment?
“It depends on what kind of company with what business model and in what market the entrepreneurs are founding,” says Retterath. “However, there are also generic red flags, such as shrinking metrics on employees, website traffic, payment data or even negative reviews on product, employee and reference platforms such as ProductHunt, Kununu or Tegus.” DDVCVs would bring all of this data together in order to To provide AI with the broadest possible information base for the pre-selection and continuous monitoring of companies.
Read too
An investor explains which startups are convincing in the crisis – and which are not
Will AI soon be the only one making decisions?
Acting in a visionary and progressive manner is part of the job description for an investor. Actually. After all, her job is about investing in exactly those products, companies and teams that drive real innovation. Venture capital is nothing more than a bet on the future.
In fact, investors are just people. Change and transformation are easier for some and not so much for others, observes pioneer André Retterath. “We see across the venture capital industry that many investment professionals are struggling with the cultural shift and use of data-driven tools and AI,” he says. “Change is easier said than done here too. In the current transition phase, those investors who proactively try out new tools and help define the software stack of a fund can be particularly successful.” And also those startups that are prepared for this.