Forensic Accounting in the Age of AI: Opportunities and Risks
October 9, 2025
While AI and other technologies can process volumes of data that once took weeks of manual effort to complete, the role forensic accountants play today is just as important as ever.
When I started as an intern in forensic accounting nearly 25 years ago, the work looked very different than it does today. On my very first investigation, I spent an entire summer sitting in a windowless room sifting through bankers’ boxes and thousands of pages of documents (by hand) in search of red flags. Today, that same review would be completed in a few minutes using technology and artificial intelligence (“AI”) to identify the red flags and other indicia of fraud.
But in the rush to adopt these new tools, we must remember that AI is not a substitute for professional skepticism, judgment, and human insight – especially when the tools themselves may facilitate new vulnerabilities, as evinced by Anthropic’s recent claims that its technology has been weaponized for cyberattacks and fraud. Meanwhile, accounting-specific risks are mounting in lieu of proper oversight: just this summer, the UK’s Financial Reporting Council (FRC) found that the country’s six largest accounting firms “do not formally monitor how automated tools and artificial intelligence impact the quality of their audits.”
In other words, while AI and other technologies can process volumes of data that once took weeks of manual effort to complete, the role forensic accountants play today is just as important as ever.
Revisiting Sarbanes-Oxley
The current focus and discussion around AI reminds me of the enactment of the Sarbanes-Oxley Act (“SOX”) in 2002. Enacted on the heels of major accounting scandals, including Enron and WorldCom, many thought SOX would be a “silver bullet” in the battle against fraud. However, in practice, it sometimes had the opposite effect – it made the approach to detecting fraud more predictable.
I remember interviewing an individual a couple of years after SOX became law. By then, SOX had been in place for a few years and should have had the effect of curtailing fraudulent behavior. Yet, during the interview when I asked him how he was able to carry on his fraudulent behavior for as long as he did without being detected, his response focused on SOX – from his view, he thought SOX was “great” because, in his words, it helped him carry on his fraud longer than he ever expected. SOX made audits more predictable, and once he understood which accounts were the focus of the audits, he shifted his activities into areas less likely to be tested.
That experience underscores a critical lesson: once a fraudster understands the rules of the game, the fraudster will adapt and exploit those rules. Frameworks designed to deter misconduct can, if they are predictable, provide perpetrators with a roadmap to hide their illicit activities.
AI risks falling into the same trap, wherein the very technologies designed to mitigate risk create new ones – whether the tools themselves create fresh opportunities for fraudsters or lead to (potentially inexplicable) accounting inaccuracies. For example, a 2024 report from the Center for Audit Quality noted that the “black box” nature of AI can make it difficult to understand how and why AI tools reach their conclusions, while others have expressed concerns that AI trained on past instances of fraud may not be able to detect irregularities designed to penetrate existing safeguards.
The Double Edged Sword of AI
Forensic accounting in the age of AI presents both opportunities and challenges. AI offers extraordinary advantages and benefits to forensic accountants:
Speed – vast data sets can be processed in minutes rather than weeks.
Pattern recognition – AI can identify patterns and anomalies that may not be discernable to the human eye.
Continuous monitoring – millions of transactions across all geographies can be monitored and “risk-scored” in real time.
Hypothesis testing – forensic accountants can test hypotheses to identify blind spots.
These capabilities allow investigations to move faster and enable forensic accountants to devote more time and energy to strategy rather than performing repetitive tasks.
But the challenges and risks of incorporating AI are equally significant:
False positives – while excellent at detecting outliers, it is the case that anomalies flagged by AI may be errors, not evidence of fraud. Without human review, false positives can derail an investigation.
Gaming the system – similar to SOX, once perpetrators understand how algorithms assign risk scores or thresholds, perpetrators can design schemes to circumvent detection.
AI-assisted fraud – fraudsters are already using AI to create synthetic transactions, falsify documents, and even generate deepfakes to obstruct investigations. For instance, Anthropic told the BBC that “North Korean operatives” used AI models to apply for remote jobs at US Fortune 500 tech companies. As technology continues to improve and become more powerful, AI-assisted fraud will become harder to detect.
Over-reliance on automation – over-reliance on AI outputs – i.e., trusting but not verifying – risks missing important nuance that comes only from context and professional skepticism. This is particularly problematic if firms, like those addressed by the FRC, do not have any metrics or key performance indicators in place to effectively measure AI’s impact on audits.
The same qualities that make AI a powerful weapon in the battle against fraud – speed, consistency, scalability – also make it dangerous if investigators accept its results without scrutinizing them.
Why AI Will Not Replace Forensic Accountants
Despite ominous predictions that AI will replace forensic accountants, those predictions are misplaced. AI changes the type of work forensic accountants do, but it does not replace the role.
AI can identify unusual activity, but only a professional can determine whether an anomaly is meaningful and should be investigated further or if the anomaly can be explained. Forensic accountants provide the critical context:
Interpreting results in light of business realities.
Weighing conflicting evidence from various sources.
Applying professional skepticism to separate genuine patterns that require investigation from noise.
Explaining findings clearly to a trier of fact.
While courts and regulators may be interested in the algorithms and methodologies used by AI, they ultimately expect a forensic accountant to explain the results. AI cannot take an oath or withstand cross-examination.
Testimony requires credibility, judgment, and the ability to translate technical findings into understandable conclusions — all of which continue to be a key role played by the forensic accountant.
The Human Element
Perhaps most importantly, AI cannot replicate the human element of investigations. A close colleague I worked with used to say “you cannot teach someone to have 30 years of experience,” and she is right. Fraud takes many forms and the reasons behind why someone commits fraud may be complex, but whatever the motive, fraud must ultimately be confronted by people, not machines.
Nowhere is this clearer than when conducting interviews. A forensic accountant draws on their years of experience to interpret verbal and non-verbal cues: body language, dialogue, reactions. The ability of the forensic accountant to adjust a line of questioning in real time can open doors that AI would never recognize. Building on this, many investigative breakthroughs come from what a subject does not say, and only a skilled forensic accountant can process those gaps in real time.
AI can help prepare interview outlines or suggest questions, but it cannot replace the forensic accountant’s experience in the moment.
Looking Ahead
As forensic accountants continue to leverage the power of AI in their investigations, forensic accountants will face new responsibilities:
Guarding against predictability: Just as SOX created frameworks that ultimately became predictable, AI is prone to a similar fate. Fraudsters are always looking to understand the rules of the game so they can exploit them. Forensic accountants must remain vigilant to this risk.
Expanding skill sets: Tomorrow’s forensic accountant will move well beyond debits and credits, blending accounting expertise with legal acumen and advanced data science capabilities.
Maintaining transparency: AI results are not self-validating. In court or before regulators, it will be the forensic accountant’s testimony — explaining how the technology was applied and what the results mean — that provides accountability.
Anticipating fraud’s evolution: Just as investigators adopt AI, so too will perpetrators. Forensic accountants must continually evolve their investigative methods to stay ahead of the fraudsters.
Conclusion
AI has already changed the way forensic accountants work – the ability to instantaneously analyze volumes of data and identify high risk transactions has been a watershed moment for the industry. This has allowed forensic accountants to shift their focus from tedious document review to spending more time on case strategy and other value added services.
But with the opportunities and benefits that come with AI, new risks emerge. The lesson from SOX cannot be ignored – fraudsters will not stop just because AI is in place. Instead, they will study algorithms, adapt their approaches, and look for ways to exploit the predictability of AI models.
Ultimately, AI will not replace the forensic accountant but will redefine the profession. The most effective forensic accountants will be those who embrace AI as a powerful tool while maintaining professional skepticism, judgment, and the human experience that technology cannot replicate. Just as SOX reshaped the compliance landscape but did not eliminate fraud, AI will continue reshaping investigations, but it cannot eliminate all fraud. The opportunity for forensic accountants rests in harnessing AI’s power while keeping human insight at the center of uncovering the truth.
(By Frank Dery, CPA. CFE, Managing Director at BRG.)
[CPA Practice Advisor]
