Before Auditoria
RingCentral is a provider of global enterprise cloud communications and collaboration solutions. For RingCentral, the finance process was complex and convoluted, with invoices in many different languages. The accounts payable team was seeking faster turnaround times for invoice extraction, and relief from doing manual invoice keying. Until recently, each team member had to open the invoice to manually capture the invoice information and hand-populate and code them into their Coupa billing system.
The AP team also had to manually work through scenarios where vendors sent many documents, in different formats and languages, including multi-page invoices, multiple attachments in one email, and duplicate invoices, all while sifting through hundreds of emails containing documents other than invoices.
Each invoice had to be attributed to one vendor record, but multiple vendors would have similar names, or no name associated in the system. Invoices needed to be coded to match a purchase order, and some invoices required specific custom fields to be included from the original invoice. Keeping track of all these nuances added to the stress and strain of invoice processing, and delayed the time it took to get the invoices into Coupa.
A final part of the manual invoice extraction process required the AP team to document that the correct information was extracted from emails and attachments with proof for the audit team to show proper compliance, adding to an already arduous and time-consuming workstream.
After Auditoria
RingCentral deployed Auditoria’s AP Invoices AI agents, pre-configured applications trained in finance and built with advanced cognitive OCR and Computer Vision technology to automatically identify invoices and extract relevant data.
All emails and activities are populated in the Auditoria console for ease of access to data and account records. AP Invoices streamlined the extraction of data from invoices received via email and email attachments, categorizing them based on the type of document and relevance, correctly identifying invoices 97%+ of the time.
The agents extracted the appropriate data fields from the document to create a draft vendor bill record, and then determined how this new bill fit with the existing data in Coupa.
The agents read invoices in various languages thanks to Auditoria’s specially built specialized language model that uses artificial intelligence designed to understand and generate human-like text but is smaller in scale and yields more accurate finance-specific results. Combined with multiple large language models (LLMs) in production as part of the Auditoria technology stack, AP Invoices delivered unprecedented human-like capabilities for the finance team.
Using Collaborative Mode, team members could review, train, and reprocess the agents to create more meaningful exception handling and recovery paths while building trust in the artificial intelligence.
AP Invoices identified the data on the invoice and mapped it to the entity selection, vendor record, PO, Master Item List entry, and other items the new record should link to, applying the appropriate coding to finalize the new vendor bill record.
The agents adeptly map extracted data to existing records in the Coupa system, creating draft vendor bill records and applying the appropriate codings. They ensure a seamless integration with the company’s existing financial systems and enhance data accuracy by classifying the confidence level of extracted data into high, medium, or low. The system learns from user corrections, progressively improving its high-confidence outputs.
Realizing Impactful Results
For the audit team review, the AP Invoices agents optimize how the AP team conducts their review and prepares for the AP and audit teams. AI agents extract the necessary invoice data from emails and attachments, documenting those that could not be processed with the right rigor around them to be reviewed and acknowledged appropriately. The audit trail provided proof to the AP and audit teams that items the invoice team dismissed in Auditoria had been correctly identified and confirmed that they were appropriately dismissed.
Auditoria’s AP Invoices have revolutionized RingCentral’s AP process by automating the complex, labor-intensive tasks associated with invoice management. The result is a more efficient, accurate, and secure process that allows the AP team to allocate their resources to more strategic tasks, significantly boosting productivity and safeguarding against potential fraud. The continuous learning capability of the system ensures ongoing improvements, making the AP process more resilient and adaptable to future needs.
Outstanding Benefits for RingCentral from Auditoria.AI’s AP Invoices
Streamlined Processing: The AI agents reduced the need for manual data entry, enabling the AP team to focus more on review and exception handling rather than routine data processing. Most invoices were processed in under two minutes, with a 96% accuracy rate.
Enhanced Accuracy and Efficiency: With real-time data parsing and creation of vendor invoice records, the AP team now handles a greater volume of invoices more accurately.
Robust Audit and Compliance: An automated audit trail of data extraction and execution activities was established. This not only helps in maintaining transparency but also ensures compliance with audit requirements. The system flags invoices based on confidence scores, with medium scores flagged for review and high scores processed directly, reducing the risk of errors.
Fraud Prevention: Enhanced email security protocols are employed to secure the shared mailbox from phishing, spoofing, and other malicious attacks. These include:
- Sender Policy Framework (SPF): Email authentication method that helps to identify the mail servers that are allowed to send email for a given domain
- DomainKeys Identified Mail (DKIM): Email authentication method that uses a digital signature to let the receiver of an email know that the message was sent and authorized by the owner of a domain
- Domain-based Message Authentication Reporting and Conformance (DMARC): Email authentication method which protects against bad actors using fake email addresses disguised to look like legitimate emails from trusted sources
