SettleSync is an AI-assisted payment reconciliation platform for merchants that our team at Bacancy Technology built to replace manual bookkeeping. It matches partial payments, refunds, and overpayments to the appropriate invoice for continuously reconciled, audit-ready books by capturing receipts from several processors, using artificial intelligence to extract data from unstructured forms, and posting transactions straight into QuickBooks Online and dispatch systems.
Event-driven pipeline connecting multiple payment processors to QuickBooks Online and dispatch systems with zero manual re-keying
AI-assisted receipt extraction (LangChain + GPT-4o mini) handles unstructured formats, validated against strict schemas
Browser automation drives a dispatch system's UI directly, despite it having no public API or direct integration support.
Deterministic settlement logic handles partial payments, multi-invoice allocation, LIFO refunds, and overpayments automatically
Every payment involved someone on staff logging into a processor dashboard or digging through an email receipt and then entering the numbers into QuickBooks manually. Mistakes could easily happen.
The client got paid by a variety of different processors, and each processor sent back receipts in their own format, with their own quirks, making it difficult to build one reconciliation process across the board.
The client's dispatch system has no public API, so posting payments to it required manual browser work, with no direct automation path available and no way to automate the process.
Applying a payment to the correct invoice, especially partial payments, multi-invoice payments, refunds, and overpayments, was a frequent source of accounting errors and audit-trail gaps.
Drawing on our fintech software development expertise, the developer built an event-driven pipeline that pulls receipts in as they arrive by email, webhook, or API and normalises them into one canonical model. This gives every payment source the same structure before reconciliation, without relying on separate workflows for each processor.
Some receipts don’t follow a clean, parseable format, so we use LangChain with OpenAI’s GPT-4o mini to extract the payment data from unstructured formats. The extracted fields are then checked against a strict Pydantic schema, so only validated data moves into the reconciliation flow.
We built deterministic matching logic to settle each transaction against the right invoice, covering partial payments, multi-invoice allocation, LIFO refund allocation, voids, and overpayments. This keeps the matching rules consistent even when a payment doesn’t follow the straightforward one-payment-to-one-invoice path.
Since the dispatch system has no public API, our Python developer built a Selenium-based browser automation suite that drives its UI directly. This lets payments flow into the dispatch system automatically and securely in real time without requiring staff to log in and enter each transaction manually.
Unified receipt capture across email, webhook, and API sources from multiple processors
AI-assisted extraction for unstructured receipt formats validated against strict schemas
Automatic matching of partial payments, refunds, voids and overpayments to invoices
Browser automation for API-less dispatch systems, eliminating manual payment entry and updates
1
October 2025 - June 2026
Eliminated manual data entry by 70% across multiple payment processors
Reduced reconciliation errors through automated, AI-assisted invoice matching
Cut bookkeeping time from 2-3 hours to under 30 minutes per batch
Automated dispatch-system updates despite no public API available
Kept books continuously audit-ready with real-time reconciliation
Unified receipt capture across 3 channels email, webhook, and API sources
| Backend | PythonDjango |
| Database | PostgreSQL |
| AI & Data Extraction | OpenAILangChain Pydantic |
| Async Processing | CeleryRedis |
| Browser Automation | Selenium |
| Integrations | QuickBooks OnlineGmail APIGoogle Cloud Pub/SubFluidPayCloverDejavooTowbook |
| Security | OAuth 2.0 HMAC-SHA256 |
| Testing | pytestautomated test suites |
| Infrastructure | DockerTraefik |
| Project Management | Jira |
Get access to an experienced team of developers and engineers from Bacancy, handpicked to ace your goals. Kickstart within 48 hours, no-risk trial.
Years of Business Experience
Happy Customers
Countries with Happy Customers
Agile Enabled Employees