In the evolving landscape of venture capital, there exists a noticeable divide between the potential of AI and the practical realities of the job. This gap highlights the intricacies of venture investing that often evade simple automation.
Evaluating a startup might appear to be a straightforward, algorithmic process. You consider the problem being addressed, assess the market size, analyze the competition, evaluate the traction, and profile the founders—all structured steps that seem ripe for automation.
When I was assigned the task of developing an AI-powered deal-sourcing tool for our investment fund, I initially believed the technical challenges would be the most daunting.
I quickly realized I was mistaken.
Through months of building and testing, I discovered that venture capital (VC) resists automation—not due to a fear of technology among investors, but because the intricacies of the role extend far beyond what meets the eye.
Here’s what I learned about the disparity between AI’s promises and the actual demands of the job.
The Data Dilemma
Let’s start with the issue of data. In the world of VC, information is often scarce, private, fragmented, and frequently outdated. For instance, a database might reflect a funding round that occurred 18 months ago, while the most recent filings from the Accounting and Corporate Regulatory Authority showcase a company’s status from the previous year.
Navigate Asia in
a new global order
Get the insights delivered to your inbox.
These data points act as snapshots of a company’s past rather than indicators of its future trajectory. The true pulse of momentum can be found in conversations with founders and reference calls, not static external databases.
This challenge is well-understood by quantitative hedge funds, who often have to devise proprietary leading indicators using publicly available market data.
In VC, you’re working with even less information, navigating more qualitative variables, and experiencing sluggish feedback loops. A misguided thesis in public markets may face consequences within weeks, while in VC, it may take years to fully recognize a poor investment.
SEE ALSO
Additionally, there’s the issue of outliers. The standout investments that define a fund’s legacy are often the very ones that fail to be recognized by predictive models. Examples like the rise of Bitcoin, the Covid-19 pandemic’s accelerative effect on certain sectors, and the transformative moment of GPT technology all serve as reminders—black swan events are not predictable.
Models cannot forecast the next significant shift based solely on data from previous trends. With the industry largely operating according to a power law, professionals rely on a scarce number of exceptional investments to offset the entire fund’s performance.
Moreover, crucial signals in early-stage investing are often qualitative rather than quantitative. The narrative a founder crafts around their project—explaining the problem, the solution, the timing, and their unique qualifications—often surpasses any numerical analysis.
A compelling story that reshapes market perceptions holds more value than any spreadsheet. In this realm, context triumphs over data, and narratives drive market movements.
The Subjectivity Challenge
Investment judgment is not merely an application of a set rubric; it’s a skill honed over years of experience with pattern recognition and valuable lessons. This judgment is inherently subjective. A deal that one partner may dismiss could be viewed as a prime opportunity by another, with both perspectives being valid.
In the development of our tool, we faced a significant choice: how much should we enforce our firm’s criteria against trusting the inherent intelligence of foundational models?
If we rigidly applied too many criteria, we risked filtering out opportunities that do not conform to established patterns—potentially the very contrarian bets that yield the best returns in VC. Conversely, over-relying on the model’s decision-making might lead us away from transparency.
Furthermore, there’s the dilemma of working with limited datasets. A fund may execute only a handful of deals over several years, which hardly constitutes a robust training set. Older investment memos may also be tied to outdated paradigms. A strategy based on software-as-a-service assumptions, for example, may lack relevance when evaluating companies rooted in AI.
The Relationship Barrier
While the issues surrounding data and subjectivity could hypothetically be addressed through advanced engineering, the relationship challenge cannot be easily resolved.
There lies a paradox: if your AI identifies a deal, so can everyone else’s AI. The moment information becomes systematized, any advantage dissipates.
The most lucrative investments in VC often arise through trust rather than raw data. Founders typically reach out to select investors before publicizing a funding round. A co-investor may reserve a spot for a fund known for its genuine value-add. A reference check hinges on the relationships that facilitate candid feedback.
These aspects are not inefficiencies waiting for disruption; they are fundamental characteristics of a market governed by asymmetric information, where access is earned, and reputation is invaluable.
Relationships facilitate access. In this industry, access constitutes the bulk of the work.
Where AI Excels
None of this suggests that AI is irrelevant in VC. In practice, it has already begun to transform various aspects of the workflow, and firms that ignore this trend risk obsolescence.
AI can efficiently scan, flag, and process opportunities. A notable application is qualifying inbound leads, ensuring intriguing companies aren’t overlooked due to time constraints.
AI is also reshaping the surrounding workflow: tools that summarize and highlight critical information in data rooms; research applications that compress market analysis timelines from days to hours; and Excel and PowerPoint agents built to reflect a firm’s unique style, aiding in initial drafts, meeting transcriptions, and outreach automation.
These enhancements are significant, freeing up valuable human resources for the most critical tasks.
The Path Forward
Since the start of this project, AI has progressed beyond simple assistance. Many tasks within the workflow are no longer augmented but automated. As cutting-edge models continue to evolve, they may eventually penetrate deeper into the realms of judgment.
And perhaps that evolution is acceptable. It may even be advantageous.
As the leaders within my firm often remind us, the act of investing is merely the beginning and simplest facet of the role. What follows—the board responsibilities, challenging discussions with founders, tough choices in down rounds, and long-term commitments to individuals—is where the true essence of the job lies.
If AI takes over sourcing, screening, and initial analysis, what remains is what has always mattered most. This may not be something to resist; perhaps it embodies the very purpose of the endeavor.
The writer is an associate, investment at Vertex Ventures South-east Asia and India.
Decoding Asia newsletter: your guide to navigating Asia in a new global order. Sign up here to get Decoding Asia newsletter. Delivered to your inbox. Free.
