Receipts strewn across the table, backs hunched over desks, eyes rapidly scanning rows of numbers and details in an expense report—this is what a typical audit department looks like in the movie, Everything Everywhere All at Once.

But auditing doesn’t always have to be that way. Technology can help automate many processes, like reviewing receipts and validating submitted data. This can reduce human error, make workflows more efficient, and give auditors more time to focus on strategic tasks.

Mark Wilfred, Director of Solutions Consulting, Southeast Asia at SAP Concur, shares how AI and machine learning can make auditing more efficient and comprehensive in the public sector.

AI for easier auditing

An AI auditing system can automatically analyse huge volumes of expense reports in a much shorter timeframe as compared to relying on manual checks, says Wilfred.

Agencies in the public sector can have hundreds or thousands of staff, and that means many receipts and expenses to audit. When auditing is manually done, only random samples of receipts or a proportion of the entire population are checked. In these cases, non-compliance – be this intentional or unintentional dishonest acts – can go undetected.

Yet hiring more staff to deal with the large volume of reports is neither a sustainable nor scalable solution, says Wilfred.

Technologies such as Concur Detect, an auditing platform powered by AI and machine learning, can help auditors to be more efficient and focused. AI can automatically detect and highlight high-risk expenses. Examples include large sums of money spent on certain categories like meals, entertainment, and transportation, which have a higher probability of non-compliance or abuse.

The system will pick out keywords such as luxury cigar brands or alcohol brands, Wilfred explains. It can also detect credit card transactions that do not align with what the staff has claimed, he adds. For example, someone can submit a claim for hotel expenses, but the card data reveals that it was a jewellery purchase instead.

The AI system can automatically sort the non-compliant transaction based on whether they have exceeded the category spend limit, as well as how confident it is that it has detected actual non-compliance. This way, auditors can easily prioritise investigating such high-risk reports.

This makes auditing more efficient, requiring less manpower and time at an organisational level. Auditors can then use their time for other tasks such as investigation, policy review, staff education and data analytics, says Wilfred.

At a foundational level, the AI supports duplicate detection and helps prevent staff from filing multiple claims for the same receipt.

Unique advantages of AI and machine learning

AI can also identify trends in expenses over time, which allows it to quickly spot any anomalies. As more records are captured, the system can leverage a large dataset to detect unusual spending patterns over time that require investigation. This would be difficult to replicate with manual auditing.

For example, a staff might have spent an abnormally large sum at a restaurant or store, which deviates from the average spending of other staff within the same organisation, explains Wilfred.

The AI system alerts the auditor to investigate this outlier and check if the expense was incurred for a legitimate business purpose. This helps the organisation detect non-compliance more easily.

Enhancing the human element

Introducing technology into the auditing process makes the tasks involved less mundane and administrative while freeing up the auditors’ time to pursue higher-value tasks like fraud investigation and data analytics.

It’s not an exciting job to keep auditing the expense reports to ensure that staff remain compliant with the organisation’s policies, Wilfred says. This leads to human resource challenges—it is difficult to find people willing to do these jobs and retain them.

AI can also provide auditors with more confidence in their checks, says Wilfred. They can use the insights generated by AI to make better decisions and drive compliance.

For instance, auditors have access to long-term data trends to identify spending patterns, which can help them get better spending visibility and identify policies that need to be reviewed.

Besides identifying non-compliance and fraud, AI and technology can also prevent further incidents from happening through education, he adds. While technology helps organisations quickly identify those who have broken the rules and their pattern of behaviour over time, each incident is an opportunity for the manager to speak with their staff and educate them on the organisation’s policies.

With AI and machine learning, auditors can look forward to a future where they no longer drown in piles of receipts, have to pore over numerous lines of data, or worry about overlooking possible non-compliance or fraud. Technology helps to automate manual tasks, returning them precious hours which they can spend on more meaningful and rewarding work.