How it works
Bank process automation, step by step
Four steps run every day, whether the source is a correspondent statement, a card settlement file or a KYC refresh queue. Underneath them sits one decision: what your team is willing to stop looking at, written down and approved.
Step 01
Connect a source
Both sides of the reconciliation arrive on a schedule: the external statement or settlement file, and the internal ledger export. Formats are recognized per source, fields are mapped once and the mapping is versioned. Duplicate file detection stops the same statement being ingested twice, and a source that changes shape raises an ingestion exception instead of quietly matching nothing.
Step 02
Match under your tolerances
The exact pass matches on normalized reference, amount and value date. The tolerance pass then works on what is left, under the value date window, amount tolerance and reference normalization rules your team agreed. Both record the rule version that produced the match, so the split between exact and tolerance matches is always visible.
Step 03
Work the exception queue
Everything unmatched becomes a worklist item: a classification, the exact difference in plain language, a suggested action, an owning queue and an age from the value date. Analysts work items rather than reports, attachments and rationale are collected on the item, and actions that move money go through maker-checker.
Step 04
Export the evidence
For a scope and a period, the export contains the reconciliation, the open items, every action with its actor and timestamp, the approvals, the attachments and the rule versions in force at the time. It is one action, because evidence assembled by hand under deadline pressure is where corners get cut.
The first eight weeks
What onboarding actually looks like
A description of the intended path, not a promise about your bank. What takes the time is agreeing tolerances and approvals, not the software.
Weeks 1 to 2
One account, both sides
Pick the account that generates the most manual work. Get the statement and the ledger export arriving on a schedule. Nothing else matters until data supply is boring.
Weeks 3 to 4
Tolerances agreed
Run the matcher, look at what does not match, and write down what each tolerance absorbs and what it could hide. Get them approved by the person who owns the control.
Weeks 5 to 6
Queues and approvals
Routing targets, roles, aging thresholds and the maker-checker matrix. This is where an operations team decides how it wants to work, and it is worth the arguments.
Weeks 7 to 8
Parallel run and evidence
Run alongside the existing process, compare the break populations, then produce a full evidence pack for a closed period and give it to whoever will have to defend it.
What we need from you
- Both sides of one reconciliation, delivered on a schedule rather than by hand.
- A named owner for the process, who can approve a tolerance.
- The routing map: who works an FX difference, who works a return, who signs off.
- Your identity provider if you want SSO from day one.
What we do not do
- Post journals into your ledger. Postings stay in your systems, made by your people.
- Make compliance determinations or file anything with an authority.
- Close an unmatched item automatically because it is small or old.
- Replace your monitoring, screening or core banking systems.
The same four steps run on every process line, and which one to start with is set out in automation in banks. The process guide behind this page is bank reconciliation automation, and the daily walkthrough for the hardest account type is the nostro reconciliation process.
Get started
Put your first reconciliation on rails
Create an account, and we will email you how onboarding works and what a first source connection looks like. The Reconciliation Break Resolver is open to try right now, on sample data, without an account.