Fraud awareness in the AI era: Human experience meets digital intelligence

As fraud patterns change with the use of AI, the question arises: Must fraud awareness evolve as well? Or has fraud awareness within a company become redundant because AI will detect all fraud or even make fraud impossible?

A study by the Lucerne University of Applied Sciences and Arts shows that confidence in one’s own ability to recognise AI-generated fraud attempts is low. Only 14.8 % of respondents believe that their company has sufficient capabilities to systematically uncover such attempts.

The potential of AI to combat economic crime is highlighted by the Association of Certified Fraud Examiners in its Anti-Fraud Technology Benchmarking Report. 32 % of participants expect to use AI as their primary fraud prevention tool - putting AI in first place. The same report showed that many organisations employ a risk-based approach to using data analytics as effectively and efficiently as possible. They focus on risk areas: payments and expenses are monitored most frequently (44 %), followed by procurement and purchasing (41 %).

We live in a time when AI is used on both sides of the fraud equation - by fraudsters and by companies seeking to protect themselves. AI enables faster and more precise detection of hidden patterns and anomalies that often elude the human eye. At the same time, it increases the complexity, effectiveness, and scalability of fraud methods.

AI is particularly effective at systematically analysing large volumes of data and flagging common fraud indicators, as shown by studies on the use of artificial intelligence and machine learning in the financial sector. Examples include:

• Reconciling expense reports with timesheets and work locations.
• Cross-referencing complex network relationships among suppliers, employees, and accounts to detect conflicts of interest, shell vendors, or repeatedly used payees.
• Identifying duplicate submissions, inconsistent information, or unusual frequencies in insurance claims.

However, people do not act solely according to rules; they are influenced by intentions, emotions, motives, and cultural and social contexts. These complex dimensions cannot be fully captured by purely data-driven or AI-based models. As a result, AI hits the limits of its capabilities when confronted with complex fraud, context-specific events, and subtle manipulations:

• Regarding unusual transactions, management overrides, or EBITDA manipulations (Earnings Before Interest, Taxes, Depreciation, and Amortisation), a critical review by experienced experts is needed to discern hidden intent and properly assess unusual fact patterns.
• Documents are often formally correct and appear flawless at first glance. To identify hidden intentions or manipulations - for example, in contracts or provisions - it is often necessary to analyse supporting documentation, communications, and internal processes in depth. Only through comprehensive consideration of context can one judge whether the fact patterns are truly appropriate and plausible.
• As AI increasingly produces high-quality forgeries that appear genuine to automated systems, digital forensics is becoming even more important. Manipulated documents are usually detectable only through detailed forensic analyses that examine various digital and physical attributes such as metadata, timestamps, and format characteristics.

Experience shows that precisely these types of fraud and misconduct lead to substantial losses and require complex investigations.

In construction, the term “keystone” refers to the central stone at the top of an arch that, through its firm placement, stabilises and holds an entire structure together. In the same way, fraud awareness is the essential keystone in combating fraud, connecting all technological, human, and cultural elements and thereby ensuring stability and resilience against fraud risks.

To make this concept more tangible, it is worth looking back at the master builders of the Middle Ages: they could not rely on modern algorithms, but built their knowledge on years of craftsmanship, a keen sense of material properties, and geometric principles. They understood functional relationships primarily through practical testing and knew exactly where and with which levers a construction’s stability could be achieved.

This historical situation is comparable to fraud awareness in the age of AI. Companies need a deep understanding of their individual risk areas and operational interdependencies to know and operate the “levers” effectively. AI can serve as a modern tool to analyse large data sets and identify initial red flags. But just as a medieval master builder could not rely on tools alone, AI today is not sufficient to see through complex, often subtle fraud mechanisms. Seasoned professionals are indispensable - their critical enquiry, intuition, and subject-matter expertise are needed to interpret the overall picture correctly and uncover hidden risks. Beyond technology and expertise, fraud awareness also encompasses a holistic risk culture that promotes transparency, communication, and responsible behaviour. As builders refined their methods over time to meet new challenges, companies must remain flexible and adaptable to keep pace with ever-evolving AI-enabled fraud schemes. Only the combination of modern technology, human judgement, and an integrity-based corporate culture provides the stability and resilience needed to withstand fraud risks in the digital age.

This article was originally published on the Economic Crime Blog of the Lucerne University of Applied Sciences and Arts.