PROJECT OVERVIEW

The AI Two-Way Match project addresses a critical pain point for accounting teams who need to validate invoices efficiently. The solution uses AI to match invoices with existing data in the Tropic system (suppliers, contracts, requests, POs, transactions) to streamline the two-way matching process that traditionally requires significant manual effort.

I was the design lead for this project, the team consisted of a Product Manager, Engineering Lead and 2 engineers.

THE Challenge

Time-Intensive Validation Process: Accounting teams spend excessive time validating bills, especially during month-end when resources are already stretched.

Information Hunting: Staff must locate contracts, agreements, and relationship owners across systems without clear connections between documents.

Entity Name Mismatches: Invoices often use legal entity names while systems use common supplier names, creating confusion.

Approval Dependencies: The process relies on relationship owners who may approve without proper inspection, creating risk.

Month-End Bottlenecks: The billing validation process coincides with other critical month-end activities, creating resource constraints.

DISCOVERY

To understand this problem better, we interviewed our internal accounting team as well as customer accounting teams. We learned that most teams were wary of introducing a new tool into their workflow, and that validating invoices was a time-consuming, manual process.

USER NEEDS

Accounts Payable Staff need:

• Quick identification of related documents

• Confidence that invoices are valid and accurate

• Time savings during busy month-end periods

• Clear insights about potential issues

Accounting Leadership needs:

• Assurance that invalid or fraudulent invoices won't be paid

• Visibility into the validation process

• Integration with existing workflows

Department Heads need:

• Reduced involvement in routine approvals

• Clear information when their input is needed

SOLUTION FOCUS

1. Minimal Disruption: The solution integrates with existing email workflows with their AP System rather than requiring new processes.

2. AI-Powered Insights: Highlights potential issues (duplicates, pricing discrepancies) to focus human attention where needed.

3. Cross-System Integration: Connects information across suppliers, contracts, requests, and transactions.

4. Clear Visualization: Presents complex relationships in an easy-to-understand format.

DESIGN

Flow chart was put together to show how AI Two-Way Match would work. Based off the flowchart, designs were created using components from Tropic’s Canopy Design System.

FINAL DESIGNS

Key elements of the solution included:

• Email-based invoice processing (forwarding invoices to a Tropic email)

• Document AI for metadata extraction from invoices

• AI-powered matching to existing Tropic objects

• Insights generation (duplicate detection, pricing and term discrepancies)

• Facilitating streamlined communication within teams

• Different flows based on AI Confidence flows: Invoice was matched to Tropic records, invoice was matched to Tropic records but need verification, invoice could not be matched.

IMPACT

While the two-way match feature hasn’t launched, its strategic intent was clear: position Tropic as a trusted source of truth for purchase order (PO) reconciliation and help finance teams streamline operations without incurring the high costs typically associated with full-scale PO systems.

Planned Impact:

• Strategic Positioning: Reinforce Tropic’s value proposition as not just a procurement tool, but a financial operations ally—helping teams enforce spend policies, track obligations, and reduce downstream reconciliation errors.

• Cost Savings: Save finance teams from having to purchase and adopt third-party PO management tools, which are not only expensive but often require heavy change management.

• Process Improvement: Improve data accuracy and trust across departments by reducing discrepancies between invoices and contract terms, while introducing greater discipline into purchase tracking and approvals.

Measurement & Adoption Strategy: We planned to measure success across several dimensions:

• Adoption rate of the matching workflow among procurement and finance users.

• Reduction in invoice discrepancies and the number of off-contract or unapproved payments.

• Time saved on reconciliation efforts as a result of better upstream data hygiene.

• User satisfaction, especially among finance stakeholders, captured via qualitative feedback and CSAT scores.

LEARNINGS

This project provided valuable insights that will inform my future work:

• Accounting Workflow Integration: I gained deep understanding of accounting teams' validation workflows and how they interact with various AP systems. This knowledge revealed that successful fintech tools must integrate seamlessly with existing processes rather than replacing them.

• Designing for AI Transparency: Finding the right balance between showing AI confidence and highlighting actionable insights proved challenging. Users needed enough information to trust the system without being overwhelmed by technical details about how the AI works.

• Email as Interface: When designing for email consumption, information hierarchy becomes even more critical. We learned to front-load critical insights and use visual cues that translate well across different email clients.

• Cross-System User Journeys: The project highlighted the importance of designing for journeys that span multiple systems (Tropic, email clients, AP systems). Users need clear wayfinding to navigate between these environments.

• Persona Expansion Strategy: Introducing a new persona requires careful consideration of how their needs align with or differ from existing users. We developed strategies for balancing specialized functionality with maintaining a cohesive product experience.