3 Challenges of Adopting AI in Banks

Updated: Jul 27


As Artificial Intelligence is getting mature, banks have been exploring the use of AI. However, there are 3 challenges they need to know;


The first one is Predictability. Nowadays, AI is driven by Neural network, which is the most popular machine learning technology. The neural network learns through a huge amount of training data.

Different data will lead to a different result. As long as training data is fed to the neural network continuously, the neural network will evolve itself and produce an unpredictable result.


The second one is Explainability. You cannot explain why you have the result from the neural network. No matter it is a correct or incorrect result.

The neural network is just a black box, with tens of thousands of values. They are responsible for producing your result. But these values make absolutely no sense, to the human being. So you cannot explain, why the result was produced.

The third one is the Training Data. As the knowledge of the neural network is built through training. The training process requires a huge amount of processed data.

In addition to the quantity, data quality also determines, the capability of the neural network. Lack of the number of training data, or poor data quality, will lead to the neural network inaccuracy. So data quantity and quality, are the most important factors in the neural network training.

These challenges are not going away in the coming few years, they will be the hurdles, of making use of AI seriously in banks. Currently, AI in banking sectors will only be used in supporting roles in business operations.

In Axisoft, we have been building a neural network to learn the meaning of financial terms. This is the foundation of many AI applications in banks, such as chatbot.


For the details of AI adoption in banks, please contact us. Thank you.

33 views0 comments

Recent Posts

See All