Machine learning has been embedded in modern day technologies for years, but the acceleration of artificial intelligence thanks to platforms like OpenAI has enabled a new rate of utilization across numerous industries.

In addition to his pioneering venture capital career, PORTAL’s founder and Managing Partner, Peter Loukianoff, had previously co-founded a data analytics company, Silicon Valley Data Science, that was later acquired by Apple in 2018. When he started PORTAL, he wanted to bring analytics innovation to the venture capital industry by applying data science for more effective deal sourcing and efficient decision making.

With his guidance, PORTAL’s lead in-house data scientist and financial engineer, Henry Keyton, built the firm’s proprietary software called PortalMatch, which assists with data-driven due diligence, predictive analytics, and risk management. PortalMatch was trained with a proprietary dataset of $211B worth of early stage transactions across 70,000 startups. Currently the model is trained to identify ventures with a strong probability of a 10X return in 5 years time with a high degree of statistical accuracy. Combining an art (deep domain expertise and relationships) and science (statistics and analytics driven by PortalMatch) approach has allowed us to quickly evaluate 100+ deals a month. With the help of PortalMatch, we can focus on doing deep diligence on 3–4 highly qualified deals out of a large pool instead of thinly spreading our resources and efforts on marginal opportunities.

“Our current predictive model is the product of roughly 12 months of effort. To inform model selection, mathematicians and financial engineers worked in concert to produce an exploratory essay on a large data set. The emphasis on research coupled with our proprietary dataset gives us an edge that most funds will not be able to replicate,” explained Keyton, who previously led a team of 20 financial engineers in an industry project for Blackstone’s Tactical Opportunities with UC Berkeley’s Master of Financial Engineering program, the number one quantitative finance program in the country. As many readers may know, Blackstone was an early adopter in integrating AI as a driver for growth and efficiency.

Keyton further elaborated, “We not only use PortalMatch as a ‘sanity check’; we use machine learning to accelerate our due diligence process. We have the ability to vet a large volume of monthly deals for which most funds do not have the capacity. With each batch of opportunities that we run through PortalMatch, we collect statistics and metrics that help us further understand the current state of certain industries and funding stages.”

So how is PORTAL exploring AI to make venture capital more efficient?

At this time, one AI tool cannot serve as a blanket solution for the entire industry. Since each fund, industry, company, and even company stage within an industry have different metrics, solutions need to be built with a specific objective in mind. However, if firms can find or build a solution to match their objectives this can save time and effort for VCs in identifying potential investment opportunities while helping them make more informed decisions. Below are several ways PORTAL is exploring AI for efficiency in its investment process:

1. Data-driven Due Diligence: AI can help automate due diligence by analyzing large amounts of data, such as financials, market trends, customer reviews, and competitive landscape, to provide insights and predictions on the potential success of a startup. 
2. Democratizing Deployment of Capital: removes unconscious bias and uncovers promising startups that are not buttressed by pedigreed investors or founders. 
3. Portfolio Management: By leveraging computing techniques, including machine learning algorithms, AI can analyze the proper mix of a portfolio and inform the investment team on risk/return correlations. Also, AI can help assess the ongoing performance of portfolio companies, track key metrics, and provide real-time insights to VC firms, allowing them to proactively identify and address any issues or opportunities to help optimize portfolio performance and maximize returns. 
4. Predictive Analytics: Through leveraging predictive analytics to forecast market trends and identifying emerging technologies, AI can predict potential exit opportunities for portfolio companies. This can help VC firms make strategic investment decisions and align their investments with market dynamics, resulting in more successful outcomes. 
5. Risk Assessment and Management: By analyzing historical data, market trends, and external factors, AI can help identify potential risks and opportunities early on, enabling VC firms to make informed risk management decisions and optimize their investment strategies. 
6. Post-investment Support: AI can provide ongoing support to portfolio companies by leveraging data analytics and automation. For example, AI-powered tools can help with talent acquisition, operational efficiency, marketing optimization, and customer engagement, thereby enhancing the performance and value of portfolio companies.

Utilizing AI can streamline the venture capital industry with data-driven decisions but there are some considerations to be aware of according to Keyton.

Emotional Attachment

“As one of my early mentors taught me, ‘venture capital is an apprenticeship business — it’s learned by doing.’ It’s very much an art,” said Loukianoff. However, the science of AI has the ability to diminish emotional entanglement from investment decisions. Machines do not form attachments to humans or ventures, whereas VCs often emotionally invest in founders or exciting products. Given that venture capital is fundamentally a relationship driven asset class, machines can provide sobering insights to investors who have been overstimulated by a charismatic entrepreneur or a provocative value proposition. AI can prevent us from “falling in love” with fundamentally flawed companies, but still allow us to take appropriate risk on potential “home runs.”

Training The Model

Leveraging AI without understanding the underlying quantitative research is a potentially dangerous strategy. VC funds that do not have the resources to retain personnel trained in empirical methods run the risk of “garbage-in-garbage-out” scenarios. It’s challenging to develop a strategy with a black box that no one in the fund understands, which means that future decisions could be compromised. Burgeoning AI tools also appear to generalize across different industries that have different behaviors. Isolating the signal from the noise becomes difficult when models are not trained with basic behavioral finance axioms in mind.

Information Privacy

More broadly utilized predictive tools could have the potential to encourage greater opacity in private markets if entrepreneurs recognize that AI is penalizing their startup for unknown reasons. Public disclosure is not compulsory for private companies, so founders with startups not favored by the algorithm can exercise the right to keep their information private.

With the advent of every technology, there exists a window of opportunity for early adopters to reap outsized benefits before the edge is essentially “arbed” out. Whether AI becomes table stakes for all future funds is certainly not set in stone. High frequency algorithmic trading funds have existed in public markets for decades, and yet their human-decision-centered counterparts have yet to succumb to extinction. At PORTAL, we use quantitative research and algorithmic development to sharpen our analytical prism, but we still maintain a human centered approach where relationships and expertise are paramount.

Want to stay up to date on cutting edge technology, founders and venture capital trends? Request an invite to our community.