Analytics, Ethical AI and the Future of Fraud Detection

Dominating and changing almost all industries, it comes to no fraud examiner’s surprise that fraudsters are using artificial intelligence (AI) to circumvent controls and revolutionize their methods of bilking people and companies out of their hard-earned money. At the 2025 ACFE Women’s Summit, Ivy Munoko led a panel discussion with Angela Kroboth, Kennedy Meda and Linda Miller on how AI and data analytics can also be used by fraud examiners to combat and deter fraud. 

At the beginning of the session, Ivy Monuko, assistant professor at Warrington College of Business at the University of Florida, asked the room of attendees how many of them had any education on AI, with almost, if not all, of the room raising their hands, she shared that not even two years ago that same question had only half of attendees raising their hands. AI innovations are coming at a sometimes-alarming speed, and it is our job to stay up to date. 

AI Like a Pro 

A common issue with large language models (LLM), like ChatGPT, is what has been referred to as “hallucinations” or when an AI model responds to a prompt with materially false information. Linda Miller, Chief Operating Officer of Audient Group, LLC, shared the importance of prompt engineering and training data. She emphasized that when we ask very nuanced questions to AI models trained on vast amounts of data, we are often times left with unsatisfactory responses. Instead, she posed, we can use specialized data from sources that we have vetted and trust, such as internal documents, in a closed LLM that only the company has access to. Then, using that model, you can ask it to find the specific answer to the question you have posed, limiting the risk of hallucinations and bias. 

Miller, when stressing the gravity and importance of these tools, said, “We’ve never lived in a world where we’ve had the ability to uncover this much information, this quickly.” 

Kennedy Meda, a fraud prevention manager and subject-matter expert for Deseret First Credit Union, highlighted the importance of educating yourself on technological use cases. She shared an industry trend she has noticed moving from a reactive approach to technology to a proactive approach, observing how many more fraud fighters are learning about and gaining degrees in data analytics.  

Bias and Fairness 

Meda delved into the ethical risks and concerns businesses have with using these models. There is a concern with protecting the data that is being used, especially with recent cyberattacks, she explained “if we are unable to protect our institution, how can protect the data we are collecting for fraud prevention?” 

Such data includes behavioral data used to create a digital fingerprint for customers, data that reflects their unique actions and information. If AI is not trained properly and uses biased data it can create ethical concerns where the model could be learning on certain genders, racial groups and demographic areas. She states, “where in reality they are not fraudsters, you just started off with biased data.” There is also the concern of fairness, ensuring the data is fair and these systems are alerting properly. “As fraud fighters it's important to know that we are not going to be replaced by AI, because AI actually needs us to make sure the data we are using isn’t biased or unfair.” 

How Do We Get Started? 

Munoko asked each of the panelists what best practices and strategies that attendees can take moving forward to implement AI and data analysis into their work.  

Angela Kroboth, an independent data analytics consultant with KonaAI and the founder of Next Step Data Insights, delved into the importance of feeling empowered to do your own research and then coming together with technology experts to communicate your needs for fraud prevention and detection tools. She stressed that by starting small and looking at the things you are already doing, you can thenbegin to implement AI models for the processes already in place. 

Additionally, Kroboth recommends going through the data lifecycle (discovery, data preparation, model planning, model building, communication of results and operationalization) to assess where improvements and operational efficiencies can be made, “Use the data analytics lifecycle and then infuse it with some machine learning and AI.”