Overview
Environment
Azure
Engagement type
PoC funded assessment
Duration
2 Months
Scope
Design and develop an AI-driven Invoice Validation solution to validate invoices against contract-defined rules
Sonata Software designed and developed an AI-powered invoice validation solution, deployed to a new dedicated Microsoft Azure environment. The solution validates vendor invoices against contract rules, detects billing anomalies, and automates compliance reporting, reducing manual review effort and financial leakage.
Customer snapshot
Industry
Healthcare
Headquarters
North Carolina
The business challenge
- Manual or inconsistent validation of invoices against contract terms leading to potential billing discrepancies
- Difficulty in identifying pricing deviations, unauthorized fees, and duplicate invoices
- Lack of automated validation for compliance with contract rules such as billing frequency, account validity, and approved businesses
- Limited visibility into historical invoice comparisons for detecting anomalies
- Constraints around working within a defined contract scope and absence of integrations with external vendor systems
Impact of the problem
- Undetected unauthorized fees (Steri-Safe, SCRF, Fuel, Environmental) and price overruns created recurring financial leakage across high-volume invoices, typically translating to an estimated 3–5% of vendor spend in similar healthcare environments.
- Manual validation was time-intensive, creating backlogs and increasing SLA risk on invoice approvals
- High dependency on manual review increased operational costs in Accounts Payable (AP), with significant time spent on low-value verification tasks instead of strategic financial analysis
PoC objective
- Validate invoices against defined contract rules
- Automatically flag discrepancies such as pricing variance, unauthorized charges, and incorrect billing
- Enable auditability and traceability through historical data storage
The Sonata solution – what we did
The hybrid AI-powered invoice validation solution automates 10 critical billing validation rules derived from the vendor contract, stores rules in Microsoft Dataverse for easy governance, and integrates with SharePoint for secure document management. A web chatbot provides user-friendly access to validation results and exception handling. The solution establishes a foundation for continuous controls monitoring within accounts payable, enabling proactive detection of discrepancies rather than reactive audits.
Validation rules implemented
- Billing site validation: Flags mismatches between invoice billing site and contract.
- Unit price, quantity and service description validation: Flags invoices missing mandatory line-item details.
- Price manipulation detection: Flags price variances exceeding 4% annual escalation from 2021 base contract.
- Unauthorized fee detection: Flags fees not listed in contract (Steri-Safe, SCRF, fuel, environmental).
- Cancelled account detection: Flags invoices with invoice dates later than account cancellation date
- Incorrect business billed: Flags invoices outside the approved business or approver list.
- Vendor account number validation: Flags missing, unrecognized, or expired account numbers.
- Billing frequency validation: Flags advance billing when contract requires post-service billing.
- Duplicate invoice detection: Flags duplicate invoices or repeated service claims using historical data.
- Emergency charge validation: Flags emergency charges that are not properly tagged.
Microsoft technologies used
Microsoft Azure
Microsoft Dataverse
Microsoft Power Platform (Power Automate, Copilot Studio)
Microsoft SharePoint
Azure AI / Copilot
Microsoft + Sonata credibility
30+ years
of partnership
Microsoft AI Business Solutions Inner Circle member
Microsoft Frontier Firm Partner
How we did it
A hybrid approach where the rules needed to address the identified use cases are stored in Dataverse, helping reduce the need for complex prompts.
Step 1
New dedicated Microsoft Azure is configured with resources, services, and access controls specifically provisioned to deploy Copilot solution. It ensures scalability, isolation, and governance aligned with development and compliance needs.Step 2
Extract and formalize the rules implemented in the POC features, along with the additional features identified for the deployment phase.Step 3
Export the solution from Sonata Power Platform and import it into New Azure Platform.Step 4
Design the data model and implement the required Dataverse tables and relationships.Step 5
Build the validation rules and logic. Refine and optimize AI models and flows to reduce complexity and improve maintainability.Step 6
Implement document injection (invoices and contracts) capability, ensuring view-only access which are stored in SharePoint folder.Step 7
Maintain historical invoices and store validation results to support reporting, auditability, and traceability.Step 8
Establish contract-to-invoice mapping.Step 9
Collect additional invoices from client to adequately cover the overall scenarios. Generate sample invoices if existing data is insufficient for full feature coverage and reject scenarios.Step 10
Conduct System Testing (ST) to ensure adequate feature coverage and validation accuracy.Step 11
Client to conduct UAT, address UAT observations, and obtain formal sign-off.Step 12
Proceed with production deployment and provide support.
Deliverables
User stories
Architecture design
Development
Test case, test execution
Results and benefits
| Metric | Improvement |
|---|---|
| Invoice validation accuracy | Rule-based validation across 10 contract rules thus improving accuracy from ~70–75% (manual) to ~95–98% (rule-based automation) |
| Financial leakage prevention | Enhanced anomaly flagging like unauthorized fees and duplicate invoices there by preventing millions in leakage annually & leading to potential savings of 3–5% of vendor spends. |
| Manual review effort | Automated compliance checks lead to reduced manual effort by ~50%, freeing finance staff for higher-value analysis. |
| SLA compliance | Faster approvals, backlog reduction by ~30–40% |
