Artificial intelligence (AI) isn’t new to the banking and financial services industry. Trades are executed by a robo-advisor; digital assistants answer queries on fixed-deposit interest rates via natural language processing; and credit scores for loan processing are tabulated within minutes by machine-learning algorithms. But as banks apply AI to make more informed and split-second business decisions, cyber criminals are increasingly using the technology to launch money-laundering scams and perpetuate fraud on consumers and financial institutions, alike. Just last year, Americans lost upward of $1.9 billion to fraud, an increase of $300 million over 2018.

Amid this activity, the industry is honing its anti-money-laundering measures. For example, Visa has invested close to $500 million in the last five years in an AI program to detect credit card fraud. Despite the growing challenges, many observers believe AI will help banks save $300 billion by 2030 by simply preventing fraud.

The question for many banks is how to cash in on the use of AI.

Better QA Is the Ticket

Proper AI deployment takes investment, time and patience. Even getting an AI engine in place takes work; banks cannot entirely sign over fraud detection to AI. Continuous human monitoring and quality checks are also required to ensure that AI stays true to its purpose, adheres to banking and privacy regulations, complies with ethical standards, steers clear of biases and continues to learn.

This calls for a dedicated team of quality engineers to train and test the AI engine, feeding it with the right data sets, identifying the correct behavioral persona and driving it toward fewer false positives. Also, in extreme cases, QA engineers need to ensure the AI does not go rogue.

A Three-Point Plan

To ensure they create an AI engine that works, we suggest that financial institutions consider the following:

  • Data: Before building an AI engine, banks need an abundance of unique data to continuously feed it. For fraud detection, banks need to think beyond the traditional rule-based data sets like login information, customer demographic and personal information. It’s crucial to incorporate less orthodox data sources such as real-time location and consumer behavior data for transactions such as a wire transfer or debit/credit card.

    For instance, in a credit card transaction, the data set should provide a linkage between size, location and device used, as well as time of transaction. This will help the AI engine create a client persona mapped with client behavior. With this information, the bank can assess each operation of a credit card transaction and quickly determine if the transaction is fraudulent. This not only reduces false positives but also helps save billions of dollars for both merchants and customers, alike.

  • Model: AI models process bulk datasets with multiple variables, which helps them identify correlations between client behavior and flag fraudulent actions that might contradict the client persona.

    For example, for wire transfers, banks can leverage smart algorithms to calculate analytic scores based on client personas built uniquely for every customer. The algorithm then traces all transactional movements and flags anomalies. Intelligent algorithms that can detect authentication fraud methodologies will advance accuracy and lower costs for quality assurance. Most importantly, these methodologies help identify bugs more quickly and accurately.

  • Process framework: A framework helps the AI engine adapt to ever-changing fraudster strategies, as well as address multiple specialized components. Each of these components performs a unique role in detecting fraud and can help with investigating suspicious activities.

    For example, quality engineers should validate the AI framework that generates transactional scores by comparing them with a fraud score assigned to an account during a fraudulent activity. This dual-score approach can drastically reduce the number of false positives as well as instances of an oversight by a fraud analyst.

Getting the Right Talent for AI

With changing IT priorities, the role of quality engineering professionals must transcend mere code testing. Because of the need for a confluence of varied skills that cut across industries and technologies, quality analysts must work not only as a gatekeeper but also as an orchestrator of quality across the lifecycle.

For banks that see AI as a means for bolstering fraud detection, a quality analyst can play the role of a data scientist, an AI model validator and a quality engineering expert with in-depth knowledge of banking norms. One of our clients, a leading financial institution in North America, was onboarding an AI engine for fraud detection. The organization turned to our team of quality analysts who were trained in data analysis, to test and train the AI engine. Together, we reduced data-driven defects by over 40% and cut delivery costs by approximately 20%.

As Banking Changes, So Does QA

With digital, banking is becoming an experience for the end user. It is much more than a transaction. This means that fewer barriers and silos, and the easier flow of data, information and money, should complement the burgeoning fintech-fueled open-API ecosystem.

With the advent of open banking, the unrelenting push toward digital and contactless payments, the emergence of newer non-banking stakeholders (fintechs offering services such as e-wallets) and third-party entities orchestrating banking experiences, more failure points are emerging that can be exploited by malicious actors.

It’s a spy vs. spy game, one in which financial institutions need to outmaneuver the bad guys to maintain the sanctity of digital banking. Precision QA can help banks make sure that the AI they apply stems fraud and money laundering, ensuring that the white hats win the day.  

For a detailed analysis, read our whitepaper “How QA Ensures Enterprise AI Initiatives Succeed.”’

For more on how quality engineering talent skills will converge to deliver quality outcomes in digital, read our joint research with Everest “The Future of Talent in Quality Assurance.”

Shantanu Chandra

Shantanu Chandra

Shantanu Chandra is a Leader for Banking & Financial Services within Cognizant’s Quality Engineering & Assurance Practice. He has over 15 years... Read more