Man or Machine? Does AI have a place in Venture Capital?

This post explores that evolution of predictive analytics and how VC's can leverage machine learning to invest in the next big thing.

Now-a-days it seems like everyone wants to be an entrepreneur. While there are hundreds of options to choose from, it’s becoming increasingly difficult for investors to find ventures that will give them 10x returns on their investment. Enter Preseries, an automated platform that helps VC’s discover, evaluate and monitor potential investments by leveraging machine learning. “Preseries works by analyzing data about startups to predict which companies have the most promise from an investor’s standpoint. It uses data provided by the companies and gleaned from public information available in databases like Crunchbase.” [1]

Preseries use machine learning as a process improvement tool to revolutionize the way VC’s assess new ventures. Many investment decisions are based on who you know and how well you present and less on how likely you are to succeed in the future. Biases in the investment process can stunt the success of a venture and inhibit the investor from making the most informed decision. Machine learning can solve that problem by collecting the available data about a startup from the internet and creating machine learning models that predict future success.

Vance Fried and Robert Hisrich explored the decision-making process in venture capital and found that “Many proposals pass through the firm-specific screen only to be rejected without extensive review … Most deals that pass through the firm-specific screen are rejected at the generic screen based upon a reading of the business plan coupled with any existing knowledge the venture capitalist may have relevant to the proposal.” [2] Preseries addresses the two most notable problems outlined in the current decision-making process, 1) the lack of time and 2) a VC’s lack of familiarity with certain industries. By leveraging data from across the internet, Preseries is able to analyze 400+ variables to make the evaluation process more efficient and less dependent on investors understanding of the space. [3]

In the short-term, Preseries is investing it’s energy into refining it’s algorithms so they can more accurately make successful predictions. Their biggest challenge is teaching the algorithm to evaluate a startup’s incremental growth and development. Kelly Nguyen, an Associate at the consulting firm BFA, noted that her portfolio company, Destcame, “has served over 500,000 customers, received over US$2M in funding, grew their team from 15 to 24 employees, and recently expanded to Mexico” [3] since they first began using  Preseries in their investment analysis, but “These achievements are not reflected in the current PreSeries score and yet they are critical pieces of information for an investor ”.  This missing information is an important indicator of Destcame’s progress and without it, new investors may not have a full understanding of Destcame’s potential.

In the next 2-10 years, as more VC firms continue to adopt this technology, it’ll be important for Preseries to provide predictive analytics on how well a company aligns with a VC’s investment theory and how their internal data is able to predict future success.

Preseries should consider integrating human feedback into its assessment as a way to provide qualitative input on a venture’s potential success. As we’ve seen with disruptive startups like Uber and Airbnb, data doesn’t always do a great job of predicting how a startup may change consumer behavior, but qualitative feedback on consumer studies can give important insights. The human feedback can come into two main forms: 1) survey and interview data about a product or a service and 2) informed input from a network of investors.

Consumer feedback data can provide insight on why a customer was willing to pay for the service or good and how the consumer believes this product stacks up against competitors. This information becomes increasingly important as startups look to scale. Likewise, this information will be important for investors as they work to understand how successful a startup may be. If Preseries was able to assist in this capacity and turn those qualitative insights into data points, they’d be become a one-stop-shop due diligence tool for VC’s.

Similarly, input from other investors is extremely important in the decision-making process for most VC’s. It’s common practice for VC’s to co-invest in deals with other firms in their network. While this helps build confidence in the success of a new venture, it may also feed into their bias even further as people in their network may share the same views or investment theories. If VC’s instead crowd sourced feedback on startups from a variety of investors with diverse experience it could help round out their opinion on a startup.

Preseries’ success won’t be realized for at least another 5 years when potential exits and IPO’s take place. In the interim, should VC’s trust the analytical judgement of a machine? Will there ever come a point in time where VC’s will be completely automated? (786)

  1. Kara Baskin, “Preseries wants to make it easier for startups to get funded,” MIT Sloan, October 26, 2017, [], accessed November 2018.
  2. Fried, V., & Hisrich, R. (1994). Toward a Model of Venture Capital Investment Decision Making. Financial Management,23(3), 28-37. Retrieved from
  3. Atakan Cetinsoy, “Machine-Learning Software That Aims to Predict Successful Startups,” Wall Street Journal, March 3, 2017, [], accessed November 2018.
  4. Kelly Nguyen, “How can investors user Machine Learning to Pick the Right Startups,” MEDIUM | BFA, December 6, 2017, [], accessed November 2018.



Make Tonight Unforgettable with HotelTonight


Machine Learning: Powering the Insurance Industry’s “Buy It Now” Button

Student comments on Man or Machine? Does AI have a place in Venture Capital?

  1. You raise an interesting point about the need to disrupt an industry that supports disruptive innovations. AI could truly reshape the way in which VCs source and evaluate deals, and could facilitate the allocation of capital into more startups. Could training data for this AI be gathered in an open source format, given the collaborative nature of VC investing? You suggest gathering consumer survey information to predict the success of a product idea; I wonder how accurate their feedback may be, as we often see customers not actually purchasing products they envision that they would be willing to. Given that technologies have been developed to evaluate qualities of job candidates who are interviewing virtually, perhaps data on more qualitative aspects of founder success could be used as well.

  2. Fascinating topic and essay.

    My concern about Preseries is whether it is able to reach a relevant size: it inherently depends on its revenue model. If they charge a fixed fee per prediction, the service may become widespread, which in turn may actually prevent it from being valuable: if everybody can read the Preseries score, is it providing any real value to the readers? Even if VCs often co-invest, they are still ultimately competing for the best deals, hence a broadly-used service has its usefulness reduced by its potential ubiquity.

    On the other hand, if the fee is variable, i.e. higher if a startup ends up being a good investment, this may be more sustainable as a business model. Assuming that VCs would compensate Preseries fairly (instead of trying to cover their use of the score), they most certainly would be willing to pay a substantial amount as long as Preseries is relevant to their scoring process.

  3. I agree with RC’s point that more qualitative aspects could perhaps be incorporated into Preseries’ algorithms to assess founder success. Particularly for pre-revenue startups, typically the key considerations VCs seek to evaluate is the founding team itself–why are they, specifically, well-positioned to launch that product/service/idea… why not another expert from that industry? Although perhaps some traits might be gleaned from their LinkedIn and articles, their soft skills might be harder to find purely based on publicly available information. And the second key consideration which I believe might be difficult for a machine to assess is around the quality of the founding team’s assumptions. But perhaps as more and more data on successful startups becomes available, these two considerations might be easier to quantify in retrospect, and reduce the human judgement and bias inherent in the current human-centric model.

  4. Very interesting use case of machine learning that you’ve brought up. I have 2 questions / thoughts that I’d like to share:
    1) Social Capital is one of the first larger VC funds that experimented with using machine learning to do deal screening and investments. It was co-founded by Chamath Palihapitiya, former Facebook Executive, and he likes to embody the contrarian voice in the VC world. Social Capital went through a major shake-up recently with the vast majority of investing partners leaving the firm (except for Chamath). Social Capital has stopped raising funds and announces that they now will only invest using their own internal money. As outsiders, we don’t have enough information to know what the root causes were, but there was speculation around internal disagreements regarding investment approach. VC appetite for adoption of machine learning to inform their own investment looks to be varied, at this stage at least.
    2) Machine learning, and especially with the recent progress in deep learning, have proven to be good at predicting trends, based on past inputs. The big assumption as discussed in our TOM class is that the relationships in the training dataset hold with the new data that we’re hoping to predict. VC is an industry whose returns is driven by its ability to look 10, 20 years into the future and predict disruption. I do believe we can all benefit from bringing more data and rigor to the investment process and thus mitigating VCs’ biases (both conscious and unconscious). However, I don’t believe the final investment decisions are something that can be entirely driven by machine learning, without human expertise.

  5. Very interesting article. To your question around the viability of AI in the long-term, I am inclined to believe that given the risk and term associated with VC-backed projects and investments, complete automation is unlikely and highly risky. I think the use of algorithmic predictions for founders’ success is an interesting proposition and of great value however, as you alluded to, heavily depends on the reliability and quantum of data. To this end, I’d suggest looking into what other VCs that have utilized machine learning are doing to improve the quality and virility of their training data. For example, Hone Capital, which is a tech-focused VC, is utilizing AngelList to analyze over 30,000 deals from the past 10 years to feed to a machine-learning model. Based on this, they have established an assessment scorecard of 20 characteristics (whittled down from a list of 400) that are most indicative of success.

  6. In answer your final question, I argue that no, there will never be a time when the role of the VC is fully automated. Particularly in the earliest stages of a venture, investors put capital into the hands of capable entrepreneurs as an expression of their belief that this person possesses the experience and personal qualities that will allow them to achieve an extraordinary outcome. At the seed stage and even Series A, decisions are not necessarily mathematically-based, but they are people-based. The questions a VC asks in final pitch meetings may not even be pre-planned–the VC reads into the body language and passion exuded by an entrepreneur to navigate the conversation. The information extracted from this encounter is not prescriptive, even if it may be guided by a VC’s own mental models. While you could argue that technology might evolve to the point where it can perform the exact functions of 1. reading nonverbal cues and 2. know the metrics associated with entrepreneurial grit & passion, part of an investor’s job is to form opinions about the future based on their past experiences (as past founders or developers of technology). A healthy VC ecosystem requires investors with a wide variety of experiences (of both successes and failures) that allows them to read into certain business and human capital contexts with their own judgment.

    I absolutely agree that to be an effective tool, Preseries will need to incorporate (find a meaningful way to quantify) qualitative data points. I argue however that at the earliest stages of venture, the qualitative outweighs the quantitative to the point that an AI-based investment recommendation would not be useful.

  7. This is a fascinating topic, and I think it’s very clear the VC industry is fully moving towards that direction. I don’t think we’ll see fully automated investing in the very near future, but all good VCs these days at the very least have some sort of data-collecting and data-monitoring strategy. In to Preseries and Social Capital (mentioned above), there are a couple of other firms you might be interested in reading about: CircleUp and Correlation Ventures. CircleUp ( pretty much makes investment and lending decisions by leveraging a massive database of consumer brands that they claim allows them to pick up signals of breakout companies before anyone else. Correlation Ventures ( offers founders investment decisions in “two weeks” by leveraging a massive historical data set of what successful deals have looked like.

  8. This is a really interesting article, thanks for sharing. One key challenge I see with this disruptive play is the feedback loop. I’d argue that the sample size and the time that it takes to realise returns (and therefore ‘prove’ success) are so long that the refinations to the algorithm referable to a particular investing context will only come into play at a time where that paradigm may no longer be in play.

    I think the fact that there are a number of well-funded players in the space, and that VC more generally is moving this way especially with discovery and initial due diligence processes suggests there is fire under the smoke, but this feels like a significant challenge that will need to be carefully addressed.

Leave a comment