SettleWise AI

Intelligent Financial Reconciliation & Autonomous Settlement Intelligence

Smarter Settlements. Stronger Business.

1. Inspiration – The Problem Behind Every Payment

Every business receives payments, generates invoices, manages outstanding balances, and tracks financial records. But the real challenge begins when these records do not match.

A payment arrives without a clear invoice reference. An invoice is partially paid. Two transactions appear similar. A customer payment remains unmatched. Finance teams spend hours investigating discrepancies that could otherwise be identified much faster.

Traditional financial dashboards show what happened, but they rarely explain why something happened or what should happen next.

We asked ourselves:

What if financial reconciliation could think, investigate, explain, and recommend actions instead of simply displaying numbers?

That question inspired SettleWise AI.

Our vision is to transform financial reconciliation from a repetitive manual task into an intelligent, transparent, and AI-assisted business workflow.

2. What It Does – Beyond Traditional Financial Dashboards

SettleWise AI is an AI-powered financial reconciliation and settlement intelligence platform that connects payment data, invoice records, and intelligent analysis into one unified experience.

It helps businesses understand their financial position, investigate discrepancies, and make informed settlement decisions.

Core Intelligence Modules

01. Intelligent Reconciliation Engine

Matches incoming PayPal transactions with business invoices using multiple attributes, including transaction amount, invoice references, dates, descriptions, and contextual similarities.

Instead of depending exclusively on exact matches, the engine evaluates potential relationships and identifies records that require further investigation.

02. Unmatched Payment Intelligence

Identifies transactions that cannot be confidently associated with an invoice and highlights the information needed for further verification.

03. Duplicate & Anomaly Detection

Identifies potentially duplicated records, unusual transaction patterns, inconsistent amounts, and suspicious financial activity for review.

04. AI Financial Investigation

Uses Gemini-powered analysis to explain why a transaction may not match an invoice, summarise discrepancies, and suggest appropriate next steps.

05. Payment Delay Intelligence

Analyses invoice due dates, outstanding amounts, and available payment history to highlight overdue invoices and potential payment delays.

06. Smart Settlement Agent

Converts financial insights into actionable recommendations. It can prepare payment-related workflows, while sensitive actions require explicit human authorization.

07. Financial Intelligence Dashboard

Provides a consolidated view of:

  • Total payment volume
  • Successfully reconciled transactions
  • Unmatched payments
  • Pending invoices
  • Outstanding balances
  • Potential anomalies
  • Reconciliation performance

08. AI Finance Assistant

Allows users to interact with their financial records using natural language and ask questions such as:

  • Why is this payment unmatched?
  • Which invoices are overdue?
  • Are there any duplicate transactions?
  • What discrepancies need immediate attention?
  • Summarise the current financial position.

3. System Architecture – How SettleWise AI Works

SettleWise AI follows a modular architecture that connects payment infrastructure, business records, intelligent analysis, and user-approved financial operations.

High-Level Architecture

          BUSINESS USER
                |
                v
      REACT + TAILWIND FRONTEND
                |
                v
        FASTAPI BACKEND
                |
       +--------+---------+
       |                  |
       v                  v
  PAYPAL API         BUSINESS DATA
  INTEGRATION        & INVOICES
       |                  |
       +--------+---------+
                |
                v
       TRANSACTION PROCESSING
                |
                v
      RECONCILIATION ENGINE
                |
       +--------+---------+
       |         |         |
       v         v         v
     MATCH    ANOMALY   UNMATCHED
   ANALYSIS   DETECTION   ANALYSIS
       |         |         |
       +---------+---------+
                |
                v
          GEMINI AI
       INTELLIGENCE LAYER
                |
                v
       FINANCIAL DECISION
            ENGINE
                |
                v
       HUMAN APPROVAL LAYER
                |
                v
        PAYPAL SANDBOX
       PAYMENT WORKFLOWS
                |
                v
       AUDIT & TRANSACTION
             HISTORY
                |
                v
       FINANCIAL DASHBOARD

Architecture Components

Frontend Layer – React + Tailwind CSS

Provides the user interface for financial analytics, invoice management, transaction investigation, reconciliation monitoring, and AI-assisted insights.

Application Layer – Python FastAPI

Acts as the central orchestration layer. It manages API requests, validates inputs, processes transactions, coordinates AI analysis, and handles payment-related operations.

Payment Integration Layer – PayPal Developer Platform

Connects SettleWise AI with PayPal Sandbox capabilities for payment-related workflows. It supports the demonstration of payment creation, approval, capture, and transaction-related operations.

Intelligence Layer – Google Gemini

Provides contextual reasoning and structured financial explanations. It assists in interpreting transaction descriptions, investigating mismatches, and generating recommendations.

Reconciliation Engine

Evaluates transaction and invoice relationships using structured matching rules and confidence-based analysis. Uncertain matches are sent for human review instead of being automatically accepted.

Data Layer – PostgreSQL / SQLite

Stores invoices, transaction records, reconciliation results, user configurations, and audit information.

Security & Approval Layer

Ensures that sensitive financial operations are validated and require appropriate user authorization. Payment credentials remain on the backend and are not exposed to the frontend.

4. How We Built It – Technology Meets Financial Intelligence

We approached SettleWise AI as a modular fintech application rather than a simple AI-powered interface.

Our development process focused on connecting independent components into one coherent financial workflow.

Technology Stack

Layer Technology Purpose
Frontend React Interactive application
Styling Tailwind CSS Responsive UI
Backend Python FastAPI API orchestration
AI Engine Google Gemini API Contextual financial intelligence
Payments PayPal REST APIs Payment workflows
Database PostgreSQL / SQLite Structured financial records
Data Visualization AG Grid & Charts Transaction analytics
Deployment Render Application hosting

The AI Workflow

The intelligence workflow follows five stages:

  1. Ingest: Collect payment and invoice records.
  2. Normalize: Standardise transaction attributes and financial information.
  3. Reconcile: Evaluate potential invoice-to-payment relationships.
  4. Investigate: Use AI to explain discrepancies and generate insights.
  5. Recommend: Present actionable settlement recommendations for user approval.

An important design decision was separating deterministic financial calculations from AI-generated explanations. Amounts, balances, and payment statuses are handled through application logic, while AI supports contextual analysis and decision assistance.

This separation helps maintain consistency and makes the system easier to validate.

5. PayPal Integration – Making AI Financially Actionable

PayPal is not simply an additional payment button in SettleWise AI. It is an important part of the platform's financial workflow.

Through the PayPal Developer Platform and Sandbox environment, we integrate payment-related capabilities into our application.

The integration is designed around:

  • Payment creation and checkout workflows.
  • Payment approval and capture.
  • Transaction information for reconciliation.
  • Linking payment activity with business invoice records.
  • Maintaining transaction references for auditability.

The AI layer interprets and analyses financial information, while PayPal provides the payment infrastructure.

This creates a clear separation between intelligent recommendations and actual financial execution.

Our principle: AI can recommend a financial action, but the user remains in control of approving it.

6. Challenges We Ran Into

Building a financial intelligence platform introduced several interesting engineering challenges.

The Matching Problem

Two records can represent the same payment even when their descriptions or reference fields differ. We needed an approach that considered multiple attributes rather than relying on exact string matching.

AI Reliability

Financial decisions require accuracy. We had to distinguish between factual calculations and AI-generated interpretations, ensuring that uncertain results were flagged for verification.

Payment Workflow Integration

Integrating PayPal into a broader reconciliation workflow required careful handling of payment states, approval steps, API responses, and transaction references.

Handling Financial Edge Cases

Partial payments, unmatched transactions, duplicate records, and overdue invoices make reconciliation more complicated than a simple one-to-one comparison.

Building a Complete Product Experience

Another challenge was bringing all these capabilities together in a clean interface that makes complex financial information understandable.

7. Accomplishments That We're Proud Of

We are proud of creating a platform that approaches financial reconciliation as an intelligent decision-making process rather than just a reporting activity.

Our key accomplishments include:

  • Designing a dedicated reconciliation engine for transaction-to-invoice matching.
  • Integrating AI-powered financial investigation into the reconciliation workflow.
  • Combining PayPal payment capabilities with contextual financial intelligence.
  • Creating a structured approach to unmatched payments and potential anomalies.
  • Introducing human-approved settlement workflows.
  • Designing a modular architecture that can support additional payment providers.
  • Bringing transaction management, financial analytics, and AI assistance into one unified experience.

Most importantly, we built around a real business problem and focused on making the solution practical, explainable, and extensible.

8. What We Learned

Building SettleWise AI helped us understand that applying AI to fintech is fundamentally different from applying AI to ordinary conversational applications.

Financial intelligence requires trustworthy data, predictable calculations, secure integrations, and transparent decisions.

We learned:

  • How payment APIs can power real-world financial workflows.
  • Why transaction matching needs multiple validation signals.
  • How AI can assist with contextual analysis without replacing deterministic financial logic.
  • Why human approval is essential for sensitive financial operations.
  • How modular architecture makes complex applications easier to develop and extend.
  • Why clear explanations and audit trails are important for financial trust.

The biggest lesson was that successful financial automation is not about removing humans from the process. It is about helping them make better decisions with less manual effort.

9. What's Next for SettleWise AI?

We see SettleWise AI evolving into a comprehensive AI-powered financial operations platform for small businesses, startups, freelancers, and growing enterprises.

Our roadmap includes:

Multi-Provider Financial Intelligence

Extending integrations beyond PayPal to support additional payment gateways and accounting platforms.

Advanced Reconciliation Models

Improving matching accuracy through historical transaction patterns, adaptive matching strategies, and feedback from verified matches.

Predictive Cash-Flow Intelligence

Helping businesses anticipate upcoming cash-flow gaps, payment delays, and outstanding financial obligations.

Automated Invoice Follow-Ups

Preparing intelligent payment reminders and follow-up workflows based on invoice status and customer payment history.

Advanced Financial Risk Analysis

Introducing more sophisticated anomaly detection and explainable risk indicators.

Agentic Financial Operations

Developing specialised AI agents that can investigate discrepancies, coordinate workflows, prepare settlement actions, and request user authorization before execution.

Business Intelligence & Reporting

Providing deeper financial insights, custom reporting, and decision-support tools for business owners.

Our Vision

We believe the future of financial operations will not be defined by dashboards that simply display transactions.

It will be defined by intelligent systems that understand financial activity, investigate problems, explain what matters, and help people take the right actions.

SettleWise AI — From Payment Records to Financial Intelligence.

Built With

Share this project:

Updates

Submission history