AI & Data Analytics: The Fraud Mitigation Dream Team

A Powerful Combination That Could Save Organizations Compliance Costs

Coauthored by Kirstie Tiernan

The advances in technology in recent years have led to an exponential increase in the volume of data collected and stored. How can investigators maximize data analytics to achieve the most effective—and efficient—results for their clients? The answer lies in harnessing the power of AI to augment the capability and capacity of an investigator. 

From paper records and rudimentary analytics tools to artificial intelligence (AI) and complex data analysis, the world of investigative technology has clearly come a long way.

To identify potential risk areas for their clients, forensic investigators have traditionally relied on limited sets of information from their clients and rudimentary analytical tools. While this may have been previously adequate, the complex business environment, management structures and data deluge in today’s organizations have given rise to unconventional data sources that add important correlations to financial data. If these correlations are ignored, there is the potential that forensic investigators may miss opportunities to mitigate instances of fraud for their clients. To further complicate the challenges faced today, this in-depth analysis, across a vast amount of data, needs to be done in a cost-effective manner.

The Problem with Traditional Analytics

The advances in technology in recent years have led to an exponential increase in the volume of data that is collected and stored. Previously, the only viable solution that was available to analyze such massive quantities of data was to take smaller, more manageable random samples from these data sets. Although this technique is certainly better than forgoing data analysis altogether, or spending months poring through terabytes of information, using random sampling of data still presents three main problems that hamper its effectiveness:

• It may not spot larger patterns of activity across a complete data set;
• Anomalous data transactions that lie outside the sample-set are missed; and
• It is a very labor-intensive process that consumes too much time and further adds to ‘analyst fatigue.’

As a result, broader patterns in the data often go undetected during sample-sets analysis simply because such patterns are only detectable if the complete data set is analyzed. When analyzing a random sample data population from a larger data set, there is a higher chance that anomalous data transactions lie outside the sample population. In addition, as ever-greater amounts of data are collected and stored, the amount of data to analyze (even with a small percentage), continually increases in relation to the volume of data, which, in turn, adds to the time, effort and cost required to perform the analysis. At best, random samples are limited in their effectiveness in providing a representative view of the complete data set.

Traditional analytics tools often require an investment of too much time, which limits their effectiveness and efficiency. Data analysis through random sampling potentially misses the big picture, risks overlooking anomalous data and takes too long to be optimally useful.

How can investigators maximize data analytics to achieve the most effective—and efficient—results for their clients? The answer lies in harnessing the power of AI to augment the capability and capacity of an investigator.

How AI Solutions Solve Traditional Data Analytics Problems

In recent years, the use of analytics or big data has had a major impact in the way businesses function. As the amount of data grows, figuring out what to do becomes increasingly difficult. The use of analytics can unlock opportunities or threats hidden in mountains of data.

Although AI technology is widely used in many industries, it is relatively new to the audit and forensic profession.

Leveraging AI to support the requirements of auditors and forensic investigators helps address each of the three main issues that have been outlined with traditional data analysis tools and processes. Using AI results in:

• The identification of related patterns of activity across the full data set;
• The detection of anomalous transactions by analyzing the complete set of data; and
• Automation, which increases speed, efficiency and confidence in the analysis.

AI platforms with machine learning (ML) continually learn from user interaction. The more exposure to data they have, the smarter they become. This is important because ML alleviates the manual maintenance and decision making that is often involved when using traditional analytic tools.

As AI systems learn more about a given data set, they can analyze secondary data and cross-correlate with hundreds of variables. As an example of how technology detects incidents that would otherwise go unnoticed, AI can be used to predict traffic and analyze weather and broad economic trends.

The MindBridge’s Ai Auditor™ leverages artificial intelligence to detect errors in financial data caused by human error or intent. The Ai Auditor automates the ingestion and analysis of data and learns from user interaction. A hybrid of ML and AI algorithms are applied against 100 percent of transactions, generating a ‘risk score’ for all transactions. Results are presented in an intuitive interface, identifying transactions warranting further investigation. By modeling an organization’s accounts, the Ai Auditor can identify all the processes that impact the finances of a business—and identify the ones vulnerable to fraud.

A few of the key ‘red flags’ the Ai Auditor identifies through data analysis include:

• Rarity: Looks at the frequency of transaction flows between accounts. The end result is an analysis of unusual flows within a ledger in conjunction with specific account flows.
• Outlier detection: Financial misstatements or fraud often go undetected because they’re covered up by other similar transactions. To address this, machine learning builds outlier detection in several layers to understand the rarity of business transactions or processes. Based on the above, an outlier score is generated based on other transactions in its group and the ledger as a whole. This supersedes a human being’s ability to identify the outliers and provides additional dimensions to the analysis—allowing the system to spot outliers that are hiding within the prescribed business processes.
• Expert systems: Built on the knowledge and industry expertise of auditors and investigators from around the world. It provides ‘expert analysis’ that is used to augment the machine’s ability to identify unusual and incorrect financial patterns.

AI systems are capable of sifting through and applying a wide variety of tests/algorithms across huge amounts of data rapidly. AI systems analyze complete data sets (not just samples) to detect errors or anomalous patterns of activity. The result is an analysis that is more complete and reduces the risk in not detecting errors or anomalous transactions.

Given the speed and accuracy of AI solutions such as the MindBridge Ai Auditor, analysts have more time available to actively work on the risks detected. This allows investigators to double-check anomalous data in the field and to resolve a problem, rather than simply sorting through endless amounts of information.

The MindBridge Ai Auditor can perform a large-scale analysis in hours, compared to traditional analytics tools that can take months to complete. The results are more comprehensive and greatly reduces an organization’s data analytics costs.

Leveraging AI-based solutions in fraud detection will help in delivering complete and efficient audits.

How to Use AI to Reduce Compliance Costs

AI-based solutions have demonstrated significant potential for cost savings in areas of regulatory compliance. While current applications of AI technologies center on digitizing and automating manual reporting and compliance procedures, AI also provides compliance teams an opportunity to automate typically burdensome tasks and enable a refocused approach on risk mitigation.

Organizations can leverage AI systems to meet their regulatory compliance by automating compliance reporting and using AI solutions to identify instances of noncompliance that require further attention. This streamlines the process by ensuring regulatory compliance much quicker than what was previously possible. It also mitigates the risk of unintentional noncompliance.

As AI systems grow increasingly more sophisticated, they will be able to tackle greater amounts of data. By leveraging AI technologies, fields like regulatory compliance, fraud detection and risk mitigation will continue to become more efficient and effective.

Kirstie Tiernan is a managing director in BDO’s Technology & Business Transformation Services practice. She can be reached at ktiernan@bdo.com.

Eli Fathi is the CEO of MindBridge Analytics Inc.™ He can be reached at eli@mindbridge.ai.