Artificial Intelligence : The new weapon to tackle financial crime risk.
When it comes to financial crime today, it appears to us as a kind of self evident after investigation that the most promising analytics tools are undoubtedly artificial intelligence. Machine intelligence constitutes a better alternative to analyst jobs because the use of Artificial Intelligence (AI) and in particular Natural Language Processing (NLP) provides a rapid capacity to identify, categorise and link entities, peoples and businesses. AI has the capacity to compile relevant information through adverse media search to effectively construct a reputational risk score. Recent advanced machine learning models such as deep learning offers best accurate results than human in financial fraud prediction, trade based money laundering, terrorism financing and money laundering detection. The development of network theory and geometric graph prediction show interesting result when we want to analyze link to link structure, name entity recognition, similarity detection and community detection.
The core arguments in favor of AI.
We will highlight 5 domains of financial crime were artificial intelligence can make the difference.
1. Reputational risk discovery and financial crime scoring.
The reputation of an entity is one of key factor to consider when referring to potential risk of an entity. AI algorithm is trained through supervised learning to effectively detect suspicious pattern of behaviour. The model analyse the semantic link dynamic, continuously to update is knowledge and predict the underlying sentiment (positive, neutral, negative) score by itself as human. That is a big advantage of financial institutions and one of new advanced technology to be complement of human intelligence activity.
2. Operational cost reduction and organisational efficiency.
Using machine learning is almost surely value for financial institution engaging in this way, it will improve considerably their compliance effectiveness and gain big competitive advantages. McKinsey & Company, reveal in its latest report that companies engaging AI and machine learning technologies in combating financial crime have been seen their compliance costs reducing by 20–30%. IBM’s study shows that using AI in the financial crime combating process gained 60 percent faster reviews and reduced investigation times by automating and switching from a manual search to unsupervised machine learning search. Compagny reduce 50 percent of rework and improve their business impact. In the same logic machine learning, increase analyst productivity and improved outcome consistency, using NLP, reduced audit effort by offering automatic real time decision making.
3. Money Laundering/Terrosrism Financing complex structure Discovery.
Rudimentary keywords search with the client’s name and a risky word such as crime or linked strings in a standard search engine is inaccurate, inefficient and create a lot of frustrating results. Looking closely, money laundering and terrorism financing appear to be a complex giant network composed of entity and individual structure around interdependence hidden relations. Understanding all this structure at human scale is one of the most cumbersome task for analysts, but not for artificial intelligence through network analysis. After ingesting search data it disentangle hidden relationship in the way to discover new money laundering typologies.
4. Customer Due Diligence (CDD) and Know Your Customer (KYC)
Screening customers before engaging in business is one of bank obligation to be compliant in regulatory purpose. Unfortunately, sanctions lists have become longer to process and requirements more complex, political exposed persons lists, terrorist financing lists, tax evasion and bribery list database are dispatching in the world and are provided in multi-languages. Review available lists to check for potential matches and take appropriate risk mitigation measures is also becoming more difficult and times consuming for analyst in charge of financial crime analysis. The old manual screening doesn’t work correctly and generate sometimes mistakes and many false positives, that is how machine learning comes into play, it improves considerably customer and payment screening using text mining and Intuitive screening. Machine learning by applying fuzzy matching algorithms in combination with natural language understanding find rapidly and accurately equivalent name, regardless its origin, geographic or cultural variations.
5. Fraud prediction and transaction monitoring.
In our complex world, fraudsters develop everyday more and more sophisticated methods, simple rules based monitoring systems such as financial amount threshold or number of transactions per day to detect cyber fraud, identity fraud, credit card fraud doesn’t work correctly. The inefficiency of rules based detection increase false positive rate and compromises bank business risk. On the way to automatically determine the best tradeoff between business risk and compliance risk, you need more advanced machine learning model to predict fraud risk. Automated fraud screening systems powered by machine learning can help banks in reducing fraud risk, algorithms are able to detect and recognize thousands of patterns on a user’s purchasing behaviour. The three arguments which explain the importance of machine learning are : fast review, the capability to analyze large and complex data sets and Efficiency.
