Accounts payable (AP) and tax teams are often overwhelmed by the sheer volume of vendor invoices they handle, which come in various formats like PDFs, images, and electronic data interchange (EDI). On top of that, they must navigate complex regulations that vary by country. Even companies with strong enterprise resource planning (ERP) systems encounter issues such as format inconsistencies and incomplete master data. These challenges can lead to delayed postings, compliance risks and missed discounts.
A transformative solution is on the rise: detect-and-suggest capability powered by artificial intelligence (AI). This approach fits smoothly into the purchase-to-pay (P2P) workflow, automating the reading and validation of invoices against existing ERP and tax rules. By offering suggestions for common issues and routing exceptions for human review, organizations can improve their operational efficiency and accuracy in reporting. This isn't just a tech upgrade; it's a strategic shift toward better straight-through processing (STP), stronger audit trails, and faster financial closes.
The challenge of traditional sales-and-use-tax processes
Accounts payable and tax teams face significant challenges that hinder efficiency. Existing processes are data dependent, and many times this data is not tax-sensitized, limiting existing automation efforts. The AI detect-and-suggest capability addresses this by reading data from source documents and supplements existing data elements to facilitate downstream tax automation.
Additionally, suppliers frequently change invoice templates, complicating data extraction and reducing the reliability of downstream controls. On top of that, different countries have their own rules for tax identification and supply locations, adding to the complexity. Manual processes only contribute to the challenges, resulting in lengthy investigations and delayed reconciliations, particularly at month-end.
The role of AI in process transformation
AI is poised to transform tax processes, especially in automating invoice validation. The detect-and-suggest capability pulls invoices from multiple sources and aligns them with master data, improving reliability. It checks invoices against ERP data to improve compliance with complex tax rules and identifies anomalies using learned patterns. This not only cuts down on exceptions but also creates a clear audit trail, improving transparency and accountability.
Implications for accounting and tax functions
Integrating AI solutions into accounting and tax functions can bring substantial benefits. For starters, AI enhances efficiency by cutting down on manual reviews, enabling teams to focus on more complex cases. With first-pass accuracy of around 95%, organizations can significantly improve productivity and refine their strategic planning. Additionally, automating invoice validation accelerates the closing process, allowing businesses to seize early payment opportunities. Compliance improves as well, with AI providing a clear control narrative that is vital for audits and overall operational effectiveness. Plus, these AI solutions can integrate with existing systems, maintaining ERP controls while increasing throughput.
A global logistics provider adopted an AI detect-and-suggest approach leveraging EY.ai for Tax built with IBM Watsonx for invoice validation, significantly boosting efficiency without the need to overhaul its core systems. This capability was integrated into existing P2P workflows, allowing for more effective invoice processing.
The solution could handle various formats — PDFs, images and EDI — directly from the P2P system. By employing advanced models to classify, extract and validate invoice data, the system achieved approximately 95% first-pass accuracy across tens of thousands of invoices. This high level of accuracy streamlined operations, enabling accounts payable, tax, and audit teams to focus on higher-value tasks and significantly reduce monthly close times.
Integration was achieved through application programming interfaces (APIs), allowing the new capability to work alongside legacy systems without disruption. Additionally, a unified data layer was created using a hybrid open data lake to harmonize inputs from multiple sources, enhancing validation consistency and identifying use cases that could potentially save tens of thousands of hours annually.
Overall, this strategy demonstrates the effectiveness of deploying a detect-and-suggest capability adjacent to the ERP system, validating and mapping data from P2P workflows, and routing exceptions — all supported by a governed data layer. This approach not only boosts operational efficiency but also prepares organizations to adapt to evolving tax and compliance requirements.]
Implementation considerations
Successful implementation starts with clean supplier master data and robust governance. To achieve that, businesses need to:
- Assign ownership and establish service-level agreements (SLAs) for supplier and tax attributes.
- Prioritize high-volume templates for supplier format mapping and define clear reason codes for exceptions.
- Integrate AI at key touch points while maintaining human oversight and compliance logging.
- Train staff on confidence scores, beginning with a "recommend only" mode and gradually moving to auto-apply low-risk suggestions; and
- Create a unified data layer to align invoice, purchase order and tax reference data across different jurisdictions.
Looking to the future
As the AI detect-and-suggest capability evolves, tax codes and vendor status will be validated earlier, right at the capture stage. This proactive approach helps finance teams analyze invoice data for patterns, making it easier to spot outliers and emerging tax risks. By integrating these capabilities with governed data, organizations can streamline continuous close practices, reduce end-of-period cleanups, and facilitate automated reconciliations.
