With the AI Checker feature in Mekari Expense, the review process for expenses such as Claims, Trips, and Purchases becomes faster and more structured. Every submission submitted by users is automatically analyzed by the AI system to identify potential risks. However, this convenience still carries potential risks, such as anomalies or indications of fraud in transactions. Therefore, it is important for Approvers and Admins to understand how the system detects risks in order to make more accurate decisions.
Through AI-based analysis, each transaction is given a risk score along with the underlying reasons. These results are then displayed in the form of warnings (warning banners) and tags throughout the approval process, so potential issues can be recognized early and addressed more securely.
The analysis results are returned to Mekari Expense in the form of scores and reasons, which are then displayed as tagging and banners on the user’s dashboard. As data accumulates, the AI system will continue to evolve to improve accuracy, especially in detecting potential fraud.
Important
Please contact us here to activate the AI Checker feature.
Information banners from AI Checker may appear on Claims, Trips, and Purchase transactions and can be viewed by Admins and Approvers during the review and approval process. This way, potential issues can be recognized early and handled more securely.
Here is how the AI Checker works.
Log in to your Mekari Expense account. Then select the “Transactions” menu if you have the Admin role or click the “Approvals” menu if you have the Approver role.
Next, click the “Claims”, “Trips”, or “Purchases” tab.
-
If there are any ‘anomalous transactions’, an information banner will appear showing the total number of transactions detected as anomalies as shown in the image below. Click “View anomalies” to continue.
Mobile users can see AI detection flags (AI detection flags) through the approval index page and transaction detail page.
The index page will then display a list of Transaction IDs detected as anomalies by AI.
-
Hover your cursor over the ‘AI’ icon on one of the transactions to display a summary of AI analysis on the detected anomaly transaction. You will immediately see the Anomaly summary. Click “Review transaction” to continue.
The AI system can detect three types of anomalies: Amount, Vendor, and Duplicate, which are classified into risk levels low, medium, and high.
You will be directed to the following view. In this case, AI detects anomalies based on the displayed risk levels.
Here is the anomaly description based on the case shown in the image above.
- In the Amount section, the detected anomaly has a high risk. In short, the claim amount is considered unusual because it is much higher than the usual average, both in the fuel and parking categories. This causes the transaction to be marked as high risk.
- In the Vendor section, the detected anomaly has a medium risk. In short, the vendor used does not match the claim category, so it is considered unusual but not strong enough to be categorized as fraud.
- In the Duplicate transaction section, the detected anomaly has a low risk. Generally, the system identifies similarities between the submitted claim and other claims previously submitted. The similarities could be the same amount, the same claim category, or close submission times.
Below are detailed explanations for each risk level and the actions that need to be taken:
| Risk Level | Description | Required Action |
| NONE | No anomalies detected. The transaction appears normal. | No action required. |
| LOW | There is a minor deviation. Possibly caused by human error or slight irregularities. | Review is recommended but not urgent. |
| MEDIUM | There is a moderate deviation. Suspicious patterns identified but not yet confirmed. | Approver needs to review and validate the transaction. |
| HIGH | There is a significant anomaly. Strong indication of fraud or misuse. | Approver must investigate before approving the transaction. |
If the transaction is indeed detected as an ‘anomaly’, you can click “Flag as anomaly” for further action.
The following notification will appear on the Claims detail page indicating that the transaction is flagged as an anomaly.
Next, the following information will appear when another approver reviews the transaction with the anomaly.
If the transaction is found not to have an anomaly, you can click “Dismiss detection”.
-
Provide a reason for the dismissal, then click “Dismiss detection”.
Information related to the transaction that has been dismissed detection, along with the given reason, will be displayed in the Status section.
The following notification will appear stating that the anomaly on the transaction has been dismissed (Detection dismissed).
That concludes the explanation about AI Checker. Learn more about the overview of the Approval menu here.