Artificial intelligence techniques provide the backbone for new analytics solutions that support modern business and decision-making processes in many different areas and industries. There are many applications in risk management:

• Chatbots to support processes in credit risk, limit application and user authorization.
• Image recognition (e.g., from social media) to improve processes in risk management, fraud detection and debt collection.
• Review and evaluation of documents or images to, e.g., verify credit collateral.
• Valuation of assets, real estate and damages using image analysis techniques.
• Continuous automatic monitoring of credit standing using structured and unstructured data.
• Advanced analysis of customer behaviour and transaction forecasting
• Early warning systems (including stress tests) based on transaction streams.

Expectations and reality

Artificial intelligence is extremely good. So good, in fact, that many people expect that its use will provide guaranteed success. They assume that AI will automatically and magically solve all their problems. This, however, is not realistic. There are many examples and situations where chatbots and AI-controlled machines made strange or even inappropriate decisions. There are also cases of AI being used for regulatory purposes, such as Black Rock, where the model had to be shelved because it was impossible to understand how it worked.

When we overcome the challenges, AI will revolutionize quality

SAS and GAPR conducted a survey on artificial intelligence in banking and risk management. The results show that AI can help in many areas, including advanced customer behaviour forecasting, credit scoring, optimization of decision-making processes and better customer segmentation. Over 80% of respondents saw clear benefits from its use, including efficiency, cost reduction and limiting the risk of human errors. They also cited benefits from the use of advanced analytics, including the use of larger data volumes and new types of data, such as unstructured data, text, images and sound. Others mentioned increased model accuracy and faster model execution, leading to a shorter time to market.

Achieving these benefits, however, means dealing with the challenges and preparing well for project implementation.

AI success factors

• Availability and quality of data. The saying “garbage in, garbage out” is definitely applicable! We need to “feed” AI algorithms with adequate quality raw data that is not artificially crafted. We also need to remember GDPR requirements and customers’ consent to use profiling data.
• Interpretability of models. AI algorithms supported by the latest technological achievements allow us to produce many models, but they must also be properly interpreted.
• People. You need a competent and creative team that knows how to apply artificial intelligence techniques in specific processes and areas. The team’s task is to develop a concept and convince sponsors of the need for investment, regardless of the level of difficulty and complexity.
• Technology. Knowledge and automation of processes.
• Cost. Investing in AI may not be cheap. But with the support of a competent team and commitment of decision makers, it will pay dividends.