B2B Reconciliation Pipeline — Live Demo
This is the running engine, not a diagram. Load the expected-settlement registry, ingest asynchronous SWIFT, M-Pesa B2B and SAP ERP feeds, run the matching pass, then remediate each exception with a full audit trail. Every record you create is written to a real PostgreSQL database and isolated to your own sandbox session.
Run it on your own exportsExpected records
0
Ingested settlements
0
Auto-matched
0
Open exceptions
0
Reconciled ledger
Nothing ingested yet. Start with step 1 to load the expected-settlement registry.
What am I looking at? (plain English)
No technical background needed — here is the same story a finance or compliance team would tell.
What problem does this solve?
Money moves through several different systems — international bank transfers, mobile money, and the company's accounting software. Each one reports payments in its own format. Teams usually check them against each other by hand in spreadsheets. This tool does that checking automatically.
What do the three buttons do?
Button 1 loads the list of payments the business was expecting. Button 2 pulls in what actually arrived from the three payment channels. Button 3 compares the two lists and flags anything that does not line up.
What counts as a problem it catches?
Four kinds: the amount is wrong, the currency is wrong, money arrived that nobody expected, or money that was expected never showed up. Each one is labelled so a person knows exactly what to chase.
What does 'Remediate' mean?
It is the fix button. It corrects the record to the agreed amount, marks it as resolved, and time-stamps who changed what — so an auditor can see the whole history later.
Is this real or a mock-up?
Real. Every click writes to a live database and the results you see are calculated on the spot, not pre-written. Nothing here is anyone's actual financial data — the figures are test figures.
Can other people see my run?
No. Your walkthrough gets its own private sandbox, and one visitor can never see another visitor's records. That same separation is what keeps patient or customer data private in regulated systems.
Why does this matter to a business?
Manual matching is slow, easy to get wrong, and expensive at audit time. Automating it cuts the review work dramatically and means every correction already has the paper trail an auditor asks for.
What does it show about the person who built it?
Design of the data flow, the database, the security boundaries, and the audit rules — end to end — with the compliance standards (HIPAA, SOC 2) built in from the start rather than bolted on.
The technical version
Asynchronous multi-channel ingest
SWIFT MT103/MT202 registry entries, M-Pesa B2B confirmations and SAP ERP documents land independently and are normalized into one canonical settlement shape before matching.
Deterministic exception engine
Matching resolves on reference, then currency, then amount tolerance. Every break is classified — amount variance, currency mismatch, orphan settlement, or expected-but-unsettled — instead of landing in a spreadsheet.
Tenant isolation and audit trail
Your walkthrough is scoped to a private sandbox session; direct table access is closed to the public and every write stamps an audit timestamp, mirroring the HIPAA and SOC 2 control boundaries used in production.
Engineers can also POST their own SWIFT, M-Pesa and SAP records to /api/public/reconciliation-demo and get a real reconciliation run back as JSON.
Want the architecture behind this — data flow, RLS model and control mapping?