In venture capital, pattern recognition is everything. Having worked in the space, first with the Oxford Seed Fund and later with AfricInvest, I observed an interesting parallel between the two funds despite being at different funding stages.

Whether evaluating pre-seed ideas or Series A+ scale-ups, it was clear certain companies sparked universal excitement, while others universal scepticism.

The “maybes” always sparked the most debate, and more often than not, the advocating investor’s eloquence and conviction determined the final decision.

These deals always left me wondering whether venture capital was truly a science and whether a world could exist where human judgement was completely eliminated from the investing process.

Human Aspect of Venture Capital

Investing, particularly in early-stage startups, is often celebrated as a blend between an art and a science. Spotting the next unicorn, recognizing the qualities of a superstar founder or betting on an industry disruptor are often framed as pattern recognition, coupled with intuition that gets developed over time.

There is undeniable value in “gut feel”, and I have personally observed investors who have developed that sixth sense to easily spot what others might miss. However, humans also struggle with complexity. When increasing variables come into play – both tangible and intangible – our decision making becomes vulnerable and biases creep in.

There is a threshold beyond which the odds of missing critical factors greatly increase for humans, while inherent biases become more controlling.

Investment firms utilizing AI to address these limitations are well on their way to removing bias and improving investor consistency.

QuantumLight

QuantumLight is an example of an investment firm that has positioned itself as the first truly systematic venture capital and growth equity firm. The firm claims to have over 10 billion data points across 70,000 VC-backed companies used to identify patterns and remove subjectivity and bias.

QuantumLight uses a proprietary AI system – Aleph – to identify venture-backable businesses and has invested in 17 companies via its data-driven approach.

While partners of QuantumLight still have to greenlight a deal, the mandate of bringing scientific precision to venture capital is at the forefront of everything they do.

Unlike most venture capital firms, the majority of the team consists of engineers building out the AI model, while investors are primarily needed for due diligence and portfolio management.

QuantumLight’s success remains to be seen but its model represents a glimpse into the future of AI investing.

Challenges?

Despite all its potential, AI-Powered investing is not a foolproof model.

Algorithmic investing relies heavily on data and for early-stage businesses, sparse or incomplete datasets could be a big limitation.

Training an AI model on the readily available historical data may skew “success” metrics towards the typical US-based SaaS/Fintech companies as those regions and industries tend to have more extensive data readily available for training.

Consequently, a model might miss out on outlier investments—overlooking potential disruptive businesses that do not match the very patterns it learned during training.

Underrepresented regions, such as Africa and emerging markets, could be overlooked if the model determines limited exits and scarce capital inflows as red flags rather than contextual nuances.

Although AI-powered investing would seek to eradicate investor bias – there is an argument for emotional intelligence such as reading founder energy or team chemistry. These qualities can not be quantified and hence incorporated into a decision-making model.

Whether you see them as blind spots or biases, there are areas where human judgement still holds an advantage over AI.

Status Quo & Anticipation

In most venture capital funds today, AI is used as a productivity enhancer.

Large language models such as ChatGPT can assist with outreach, workflows and high-level market research. CRMs such as Affinity help with relationship management and notetakers such as Otter.ai and Fireflies.ai are regularly present during investor-founder intro calls.

Many of these tools are utilized prior to deal-closure at the sourcing and due diligence stage but ultimately feed into a decision eventually made by a human.

Outside of decision making, elements of the VC process remain difficult to automate; portfolio management, coaching founders through layoffs, being available as a late night sounding board are all important parts of the wider venture capital puzzle that still require a human touch.

Decision making may be the next frontier for AI deployment and platforms like QuantumLight’s Aleph hint at a future where AI not only guides decision making but actively drives outcomes.

Whether this shift leads to better outcomes remains to be seen.

Science But Not Fully Yet

It’s clear that AI can process information at a scale and speed no human ever could. Still, when it comes to all investment decisions, some elements will always be difficult, if not impossible, to quantify.

Emotional intelligence, gut feel, founder chemistry, and the ability to read subtle cues in a pitch meeting or a negotiation all factor into the judgment call of backing a team. These are dimensions of investing that still rely heavily on human intuition.

We likely remain far from a world where automation covers every stage of the pipeline from sourcing to exit but a well-designed model can certainly broaden what a VC fund considers and improve the consistency of decision making.

The very essence of a startup is to break patterns and deliver outsized returns to investors by doing what has never been done before. As models are trained to excel at pattern recognition, the fundamental tension between familiar success markers and qualities of true disruption will be at the core of algorithmic investing.

AI will undoubtedly reshape the investment process but as we wait to see the returns of algorithmic investment decisions, the final judgement on those “maybes” still belongs to the human in the room.