Thoughtful Automation: The Development of Verify

Organizations lose an estimated 5% of revenue to fraud each year, according to the Association of Certified Fraud Examiners (ACFE). Granted, not every questionable expense is associated with intentional fraud: 89% of business travelers said that they submitted at least one travel expense that they weren’t completely sure complied with their company’s travel policies in a 2022 SAP Concur pulse survey. But having processes and auditors in place to identify out-of-policy spending is critical to protecting a company’s bottom line in either scenario.

More than a decade ago, we introduced a pair of services called Policy Audit and Receipt Audit to supplement customers’ finance departments with a team of auditors employed by SAP Concur. Customers could purchase a set number of audits, and our team would review those reports for errors, potential fraud, or anything that might conflict with company policy.

As technology evolved, we started building AI models that replicated the checks being performed by our team. The purpose was not to replace the team but rather to supplement them, by increasing the scale of reports that can be reviewed and dedicating our auditors’ time to the most challenging or problematic reports. We introduced this functionality to our customers, which eventually became the Intelligent Audit solution that is available today.

Intelligent Audit has helped expand the number of reports that customers are able to audit. Before this solution, if a customer paid for 20% of their reports to be audited by SAP Concur, 80% of their reports would need to be quickly reviewed by a manager with no thorough compliance checks. Using AI enabled our auditing team to focus on specific reports with potential issues, not just random samples.

While many of our customers find value in tapping the SAP Concur team of auditors to review their reports, others are more interested in using our AI-powered solution to supplement their in-house team of auditors. To best address their needs, we developed a user interface for our AI-based auditing software, called Verify.

When customers opt in to Verify, the solution can audit 100% of their expense reports based on a configurable list of receipt- and policy-based rules. This list is derived from best practices we learned through offering Concur Audit Services, and it helps address some of the most common auditing challenges our customers experience. Over time, we’ve continued to improve accuracy of audit rules through training the AI behind Verify.

In strengthening the AI, we also bear in mind a confidence metric called operational bypass. We define operational bypass as the technology’s ability to answer a question without human intervention. For anything that the AI is unsure about, we augment the technology with a 24/7 team of SAP Concur auditors to confirm accuracy before serving reports to the customer. If the AI can confidently complete an audit check, the human is not needed to step in and validate. Verify is a great example of how humans plus AI technology can deliver better results than either could independently.

Our early development focus for Verify leveraged English language receipts, which represent a significant data set for the development and training of the initial AI model. Our next priority is applying Verify for non-English languages. That said, our technology is not starting from zero in other languages. We have receipts coming into our solution from around the world, already helping to train the AI to improve operational bypass and launch in different languages and markets.

Nearly a third (32%) of occupational frauds are associated with a lack of internal controls—and the introduction of controls is connected to lower fraud losses as well as quicker detection. Solutions like Intelligent Audit and Verify provide those internal controls necessary to drive compliance and identify potential fraud, while helping to minimize team workloads and reimburse employees faster.

Learn more about Intelligent Audit and Verify.