Introduction
Accounts Payable (AP) teams often handle invoices received through various channels including email syncs to Finance Operations, Supplier portals where invoices are manually downloaded, and direct submissions to platforms like the Coupa Supplier Portal (CSP). Regardless of their origin—whether from email, supplier portals, or service desks—all invoices funnel into ServiceNow, setting the stage for automation to take over.
The Challenge
Invoices need to be entered daily into the Coupa portal, a task that involves meticulous attention to detail and adherence to numerous rules. These include distinguishing between PO-backed and non-PO invoices, verifying supplier details, ensuring correct currency usage, and checking for discrepancies in the total amounts. This manual process is not only labor-intensive but also prone to errors that can significantly impact financial operations.
Innovative Solution with a Human Touch
Our solution involves two bots designed to optimize invoice processing while incorporating essential human oversight:
Bot1
Utilizes UiPath’s Document Understanding Invoices Model to extract data from invoices. Given that no extraction is 100% error-free, this initial step integrates a crucial human-in-the-loop system. Extracted data is presented to users via UiPath’s Action Center, where it awaits human validation. This process ensures that even if the AI makes an initial error, it doesn’t propagate through the financial systems unchecked.
Human Validation
If the data is accurate, users approve it directly within the Action Center, triggering Bot2. If discrepancies are noted, users can make corrections on the spot. These adjustments are vital not only for current accuracy but also for teaching the AI model to reduce future errors, thus closing the machine learning feedback loop.
Bot2
Once data is validated, this bot inputs the corrected and approved details into Coupa, streamlining what was once a cumbersome manual entry process.
Approach to Enhancing Accuracy
To increase the reliability of our AI model, we specifically trained the UiPath invoices model on the top 50 US vendors’ invoices. This targeted approach has significantly enhanced accuracy, ensuring that the AI’s performance improves continually as it learns from real-world data and human corrections.
Benefits of AI-Augmented Invoice Processing
Efficiency
Automating the data entry process significantly reduces the time AP teams spend on manual data entry
Accuracy
Human in the loop ensures that the AI’s outputs are verified, greatly reducing the risk of financial discrepancies
Scalability
With bots handling routine tasks, the AP team can manage greater volumes of invoices without additional strain
Continuous Improvement
Each correction feeds back into the system, continuously improving the AI model’s accuracy over time
Conclusion
By integrating UiPath’s advanced AI with a strategic human-in-the-loop approach, our invoice processing solution not only optimizes efficiency but also harnesses human expertise to ensure accuracy. This powerful combination allows financial teams to focus on more strategic tasks, secure in the knowledge that their operational workflows are both swift and precise.