Bringing AI-driven decision making to startup investing is a hotly-debated topic. Conversations quickly turn to startup investing being an artisan trade that AI can’t capture the nuance of, and concerns are raised that algorithms will exacerbate known biases in the system.
Much of the solution benefits often get overlooked.
So we want to be transparent about our views in this area, the work we are doing, and where we know we need to improve.
Two beliefs drive our thinking:
This first is that human intuition is a far from ideal decision making instrument
Our brains are inflicted with many cognitive biases that impair our judgement. This is the result of hundreds of thousands of years of evolution. Humans developed a system of reasoning that relied on simple heuristics, processing less information and causing impulsive decision-making to become prevalent in the descendant population.
As humans, we give more weight to vivid or recent events, group subjects into stereotypes, and feel compelled to find intelligent explanations for what can be just random noise.
Therefore, relying heavily on human intuition when it comes to decision making leads to suboptimal decisions. This includes startup investment decisions too, where the numbers tell a story; 95% of VCs are not generating the required returns, more than 50% of investments lose money, and $70bn+ is invested annually in startups that fail.
95% of VCs are not generating the required returns, more than 50% of investments lose money, and $70bn+ is invested annually in startups that fail.
Our role is not to blame the investor. They are working hard with the tools available to them. But there is just too much information for a human to process — too many startups to analyse, new technologies to stay on top of, industries to understand, and niche markets to be aware of. And the pace of change is only accelerating.
Human investors tend to get hung up on very few data points, struggle to think about all the relationships between the data, and prefer simple linear relationships.
In contrast, algorithms can take into account millions of data points which go beyond the reach and network of any individual investor.
Our second belief is that it is not yet possible to predict startup success from the data that exists in the public (and paid for private) domain
If your algorithms need to predict the future, then you need excellent sequential data. And to put it simply the data (or to use the correct term — information) in the public domain is not good enough. It’s not rich enough, contextualised enough, structured or tagged properly to support proper predictions.
That’s why today our use of ML and AI in startup analysis is not about trying to predict startup success. We capture a unique data set, without requiring direct input from the entrepreneur, next to many public and paid for data integrations and feed this into our models. Our goal is to score (and eventually price) the potential risks around an investment to help investors analyse more deals, support better allocation of capital, and to help the right startups get funded as early as possible.
Our goal is to score (and eventually price) the potential risks around an investment to help investors analyse more deals, support better allocation of capital, and to help the right startups get funded as early as possible.
Our algorithms will not be perfect from the start and we accept that they will initially not be free from bias or errors — there is simply no way of getting around that. However, the only way to uncover these problems is to first be able to properly record, audit, and understand it.
Once our data starts to mature (in 1–2 years) we will use ML to optimise our algorithms and find segments in our data that best explain variance at fine-grain levels, and crucially even if they are unintuitive to our human perceptions.
AI has no problem dealing with thousands or even millions of groupings. And AI is more than comfortable working with nonlinear relationships, be they exponential, power laws, geometric series, binomial distributions, or otherwise.
We can then genuinely understand how much bias and errors we had unfortunately coded, and be in a position to eradicate it.
To accelerate our improvement over the coming months we will work with AI ethicists and academia to review and ultimately publish our work.
No investor in the world currently does this, or is able to. No investor who looks back over 2, 5 or 10 years and analyses every startup investment they did (and did not) make, and then track how this played out. So when challenging bringing AI-driven decisions into startup investing we encourage people to think about the wider benefits compared to status quo and its known imperfections.
Summary
When challenging AI-driven decision making in startup investing, we like to draw comparisons with the argument that driverless cars are not safe — when we know overall they are significantly safer than humans behind the wheel.
Using a machine that is not limited by network, knowledge, or capacity to analyse makes sense. To be clear, we are not talking about removing humans from the loop. Rather we want to elevate humans to focus on the work where their value add is stronger. For example, developing investment theses, final investment selection and deal negotiation, and fostering relationships to accelerate the startups growth.
The value of AI is making better decisions than what humans alone can do. When it comes to startup investing, this can create a step-change improvement in investment efficiency and quality.
The best investors over the next decade will be those who find ways to fully leverage the value contained in their data and bring AI into the workflow as a primary processor of data.
—
About Koble
At Koble, we’re building No-Code ML infrastructure for private market investors.
Investors easily create personalised models that deliver competitive advantages on top of Koble infrastructure. In weeks, with no code.
Our AI and ML powered platform automatically scores startups across Team, Market, and Traction. Today we help investors quickly and thoroughly source, analyse, and score thousands of startups.
Koble’s algorithms will calculate the probability of the startup raising future investments and becoming an outlier. Ultimately, Koble will surpass human capability and revolutionise startup investing.
Learn more at koble.ai
