With AI, algorithms are able to perform tasks much better, faster and with less errors than humans. A recent research concludes in a more subtle way: under certain conditions, humans perform better than machines when it comes to decision making. In which context humans beat the machines? What are the implications for strategy and resources allocation?
Does an algorithm perform better than Business Angels to select startup to invest in?
In their research, Blohm & al. compare the investment returns of an algorithm with those of 255 Business Angels investing via an angel investment platform. In particular, they explore the influence of 3 human biases and experience on business angels returns:
- local bias (favoring companies located in the same geographic area),
- overconfidence bias (tendency to overestimate their knowledge and investment skills),
- loss aversion (tendency to be more sensitive to potential losses than to potential gains).
The results show that business angels with high level of decision biases have lower investment returns than those who show lower level of biases. Additionnally, results show that the algorithm outperforms the majority of business angels.
But there is one category of investors which perform better than the algorithm: investors who manage to use their experience to suppress biases in their decision making are able to produce better returns than the algorithm.
This study shows that even in high uncertainty contexts such as angel investing, algorithms perform better than experienced humans. This has an implication on the strategic importance of such algorithms.
As the scope of activities which are under threat of being automated is important, companies which built an advantage out of their human talent and processes see this advantage vanishing and the one of companies using algorithm to make decisions increasing. But, as the information to train an algorithm for making a specific decision is widely available, as well as the talent, software and infrastructure, the uniqueness and disctinctiveness of such algorithm reduces. Taken together, we may expect some commoditization of automated decision making: more available and easy to access, then less disctinctive and source of competitive advantage (unless the company has proprietary access to a very specific and valuable dataset to train the model).
The study also concludes that a category of human stil perform better than the algorithm, the one able to reduce their biases with the accumulated experience. This has an implication of the strategic importance of highly experienced human resources.
As the type of decisions and contexts for which algorithm perform better is more important, in order to make a difference, companies will need to invest a lot of resources to attract, retain and upskill experienced managers.
Automated decision making is likely to become a commodity and the source of competitive advantage still is human based but limited to the most experienced resources.