Healthcare organizations process enormous invoice volumes, but non-PO invoices remain one of the most difficult areas to automate. Unlike PO-backed invoices, there is no originating document, no structured approval chain, and often inconsistent supplier information. That complexity has historically forced AP teams into highly manual processes.
At BLUEPRINT 4D 2026, Jeff Dudas shared how Sarasota Memorial Health Care tackled that challenge head-on by building an AI-driven automation framework for non-PO invoice processing inside PeopleSoft.
The result was not just a technical success story, but a practical blueprint for organizations looking to modernize AP operations without massive consulting teams or expensive custom software.
Why Non-PO Invoice Automation Is So Difficult
Jeff explained that many AP automation tools are optimized for PO invoices because those transactions already have structured data behind them. A PO invoice has a supplier, requisition, approval history, and matching logic already established in the ERP. Non-PO invoices are far more chaotic.
At Sarasota Memorial, non-PO invoices included transportation vendors, utilities, consulting services, food services, and legal invoices. Many came from small vendors with inconsistent formats and naming conventions.
One transportation invoice example highlighted the issue perfectly. The invoice contained a company name that did not match the supplier master record, requiring the system to intelligently determine the correct supplier before any accounting could even begin.
That complexity is one reason many third-party AP automation tools struggle with non-PO processing. Organizations often end up rebuilding PeopleSoft business logic inside another platform, including duplicate checks, chartfield validation, supplier lookups, and approval routing. Jeff and his team intentionally wanted to avoid that trap.
Starting With the Real Business Problem
Sarasota Memorial processes more than 300,000 invoices annually, including approximately 35,000 payment requests per year.
When the team analyzed the workload more closely, they discovered something important: a relatively small number of employees were entering very large volumes of repetitive non-PO invoices with nearly identical accounting structures.
That insight shaped the entire automation strategy.
Rather than trying to fully replace users, the team focused on eliminating repetitive manual data entry while preserving visibility and control. The invoice dates, invoice IDs, and amounts changed regularly, but much of the accounting and routing logic stayed consistent.
That made the process an ideal candidate for AI-assisted automation.
Lessons From Early Automation Attempts
This was not Sarasota Memorial’s first attempt at AP automation. Jeff walked through multiple earlier approaches that helped shape the final solution.
The first generation relied heavily on robotic process automation (RPA). The organization used scripts and bots to scrape invoice data, log into PeopleSoft, and create vouchers automatically.
While functional, the approach proved fragile. Minor UI changes, caching updates, or prompt table behavior could break the automation. Inconsistent accounting data from source systems also created failures that required manual intervention.
A second-generation solution introduced OCR and AI technologies that improved invoice extraction, but still required extensive vendor-by-vendor training. Every new invoice format had to be manually trained before automation could work reliably.
That became a scalability problem.
Jeff described reaching a point where his team simply stopped onboarding new suppliers because the training overhead was too high. The technology worked, but operationally it was not sustainable.
The Turning Point: Generative AI Without Training
The breakthrough came when the team tested modern generative AI tools against a collection of completely unfamiliar invoices.
Instead of training models repeatedly, Sarasota Memorial submitted roughly 30 random invoices and receipts to a GenAI-based invoice extraction process. The AI successfully identified invoice details and returned structured XML data without prior training.
That changed everything.
The organization realized it could finally process invoices dynamically instead of maintaining endless training libraries for every supplier layout.
The architecture they built combined several technologies:
- PeopleSoft
- Integration Broker
- UiPath
- Generative AI
- OpenSearch
- Custom PeopleTools development
Importantly, they deliberately kept core business logic inside PeopleSoft. Duplicate checking, supplier validation, approvals, and accounting rules all remained within the ERP rather than being outsourced to external AI tools.
That decision simplified governance and gave the team full control over future changes.
Building AIPRO: A New Invoice Processing Framework
To support the new model, Sarasota Memorial created an entirely new transaction framework called AIPRO (AI Invoice Processing).
Rather than pushing invoices directly into vouchers, AIPRO functions as a staging and validation layer between invoice intake and PeopleSoft Payment Requests.
This design solved several operational problems:
- Users retained visibility into invoice status
- Errors could be corrected before voucher creation
- Transactions could be tracked end-to-end
- Approval routing remained inside PeopleSoft
- Users no longer had to monitor third-party systems
The team also built a streamlined invoice upload interface directly inside PeopleSoft. Users simply upload invoices through a lightweight attachment page that tracks submission status and processing progress.
One of the most innovative aspects of the solution was the side-by-side invoice review interface. Instead of forcing users to open attachments separately, AIPRO displays the invoice image alongside extracted invoice data on the same page.
This significantly improved usability while simplifying validation and corrections.
Solving Supplier Matching With OpenSearch
Supplier identification was one of the hardest problems the team faced.
Initially, they maintained static lookup tables that mapped extracted invoice names to supplier IDs. Over time, Jeff grew frustrated with the maintenance burden and searched for a more intelligent approach.
The answer came through OpenSearch integrated with PeopleSoft search capabilities.
The extracted supplier name is passed through OpenSearch, which compares it against supplier records and returns likely matches with remarkable accuracy. According to Jeff, the system identifies the correct supplier roughly 98–99% of the time.
When matches fail, users can manually correct the supplier and save accounting defaults that improve future automation.
This learning model allows the solution to become smarter over time without requiring extensive AI retraining.
Keeping Humans in the Loop
One of the most important themes throughout the session was the idea that automation should reduce work, not eliminate accountability.
Earlier automation efforts created “black box” experiences where users lost visibility into whether invoices had processed successfully.
The AIPRO design intentionally keeps users engaged through review, validation, workflow, and approval processes. Rather than manually entering invoices line by line, employees now focus on verification and exception handling.
That shift improved both efficiency and user confidence.
Jeff emphasized that organizations should avoid creating new monitoring burdens in the name of automation. A successful AI implementation should eliminate tasks, not create additional systems users must constantly watch.
A Practical AI Strategy for PeopleSoft Customers
Perhaps the most impressive aspect of the project was how lean the implementation remained.
Sarasota Memorial’s PeopleSoft support team consists of only five people supporting FSCM, infrastructure, development, upgrades, and projects.
The AIPRO solution was developed internally with no consulting costs and no new software purchases. Instead, the team leveraged enterprise tools that already existed within the organization.
That philosophy shaped the entire project.
Rather than chasing flashy AI features, the team focused on practical business outcomes, careful prototyping, and using each technology for its strengths.
For PeopleSoft customers exploring AI initiatives, the session offered an important reminder: modernization does not always require replacing core systems. With thoughtful architecture and targeted use of AI, organizations can dramatically improve operational efficiency while continuing to leverage the strengths of PeopleSoft.
Want more?
Explore more content and resources to help you get the most from your Oracle investment:
- Visit the Quest Learn Library for blogs, how-to demos, and on-demand sessions.
- Connect with peers in one of our Quest PeopleSoft Community User Groups to swap stories, ask questions, and share tips with other users facing the same challenges.
Not a Quest member yet? Join today and tap into the ultimate Oracle customer network.