8 ways for banks to use AI


If a bank gets in touch with a software company and wants to know how it could use artificial intelligence (AI), that means two things. First, the bank is willing to consider AI, and second, that it really does not know much about the subject. Banks want AI, and they have the necessary budget. However, they do not necessarily know what they actually want to do. This is the big question that is currently exercising the banking industry: We know AI can help us, but how? 

I think the answer may be simpler than banks realise. The best way to come up with answers is to look at how other industries benefit from AI and transfer some of their ideas. When we look carefully, there are plenty of exciting possibilities on the horizon. That sounds fairly lighthearted, but genuinely, the best way to find is to start looking. We saw with the use of big data how sectors could learn from each other’s experience, and it’s the same with AI. I think there are eight main answers to the question of how banks can use AI.

1. International transfers 

Text analytics methods can alert you to critical or dubious cases using nested text fragments from transaction logs. You can use AI to identify patterns and anomalies. The procedures must be intelligent, i.e., self-learning, so that the system learns how to separate the bad from the good.

2. Money laundering 

Banks fear fines for authorised transactions that turn out to be criminal. Investigators therefore compete to individually examine suspicious transactions. However, AI can be used to reduce the number of false positives via money laundering optimisation procedures. With Gradient Boosting (a form of AI), it is possible to reduce costs by around 20% quite quickly.

3. Call centre feedback 

Agents and customers may not get along on the phone. This can cost the bank a lot of money because supervisors and consultants have to move in to manage the process, and it is not easy to optimise. It is now possible to convert the spoken words into text and put them into a self-learning algorithm. This may eventually replace the agent’s intuition and suggest much better solutions for the customers.

4. Expert opinion 

Corporate loans are decided based on expert opinions. These are very complex, and eventually come down to a decision by a single person, albeit drawing on pages and pages of detailed reports from others. The decision may be yes or no, and may also turn out to be right or wrong. Much depends on whether the correct material has been pulled together in the report, but also on the weight placed on each piece of information by the decision maker. AI can help by identifying patterns that can provide useful information to support a sensible long-term decision.

5. Enriching the advisor's knowledge

Of course, consultants know best about their customers. They know about them from personal conversation and contact. They may not, however, know everything about them, especially if the client is mainly active in a different geographical area. There are, however, text recognition methods that can provide information from a range of media and other online sources, providing a personal dashboard that may be business-critical for consultants. AI is useful because, again, learning from experience helps to improve the applicability of the dashboard.

6. Lending 

The more you know about a loan applicant, the more confident you can be that the loan will be paid off. Classically, banks look at the history of the customer and their credit information. But are these sources of information really enough? What if geolocation could also use residency information to get more clarity? What if AI procedures can find out where the most reliable customers live or work? This would not replace existing procedures, but certainly might augment them.

7. Online banking 

We can use AI to automate regular interactions between the bank and its customers. This might mean, for example, that customers no longer have to initiate regular transfers. Care will be needed, but the AI procedure is happy to take over in limited circumstances.

8. Trading 

Traders can sometimes make 300 trades a day. AI can put a robot beside the trader, which can suggest actions 10 minutes in advance, improving decisions. The robot does that using AI algorithms, depicting human intuition mathematically.

To be honest, only examples one to six are genuinely realistic. Seven and eight are fantasies. They may, however, not remain fantasies forever. After all, every innovation begins with a fantastic idea that at first seems impossible. These ideas actually came from developments in the military and defense ministries, who pioneered email and GPS – and we all know what a success those have turned out to be! My guess is that the military is already working on the technologies of tomorrow. It makes sense to learn from them in the way they are thinking about applying AI. 

Traders can sometimes make 300 trades a day. #AI can put a robot beside the trader, which can suggest actions 10 minutes in advance, improving decisions Click To Tweet

About Author

Christian Engel

Based in Germany, Christian Engel is a Head of Banking experts and advisors for SAS

Christian Engel has lead a group of strategic business analytics advisors for key SAS accounts since 2006. His academic background is in mathematics and he completed his Diploma degree with concentrations in Operations Research in 1996 in Darmstadt. His day-to-day work involves calculating the value contribution of analytics software, optimizing analytic platforms for departments, and innovation projects related to new software technologies.

Leave A Reply

Back to Top