Reconciliation at scale for payments teams
Know exactly why your books don't balance.
AnselIQ matches financial data automatically, down to every record, millions of them in minutes, and its AI suggests the matches your rules miss. When something doesn't tie out, AnselIQ shows you the difference, the movements behind it, and the fix, with every step recorded.
Open source under the Apache 2.0 license. Run it on your own infrastructure.
Cash balance
$1,284.50
Unreconciled
$1,940.00
14 movements
In progress
-$15.50
5 reconciliations
Mismatched
-$640.00
4 reconciliations
Cash balance = unreconciled + in progress + mismatched. Final reconciliations net to zero.
The problem
Reconciliation breaks quietly.
A balance with no explanation
The clearing account is off by a few hundred dollars and nobody can say which payouts, fees or deposits make up the gap.
Exceptions without context
A mismatch lands in a spreadsheet or a ticket, separated from the transactions that caused it and from whoever is meant to fix it.
Fixes with no trail
Write-offs and corrections get edited in place, so months later no one can show what changed, when, or why.
How it works
From raw files to a balance you can explain.
-
01
Ingest every source
Files are picked up automatically as they arrive, in CSV, BAI2, JSON or fixed-width format, and recorded on a double-entry ledger.
-
02
Match with rules and AI
Define how two sides relate, one-to-one or one-to-many, and which fields and dates must agree, including offsets like "settles two days later." What the rules miss, AI suggests for your team to approve.
-
03
Explain every difference
Every balance breaks down into unreconciled, in-progress and mismatched amounts. Each mismatch shows what came in, what went out, and what it matched on.
-
04
Resolve with an auditable record
Write off, unmatch or wait for the missing movement. Each fix posts a ledger entry, closes its exception automatically, and can itself be reversed.
Mismatches, explained
The answer is on the record, not buried in a spreadsheet.
When a payout and its bank deposit don't agree, AnselIQ shows the two totals, the difference, and the exact value they were matched on. The exception that tracks it sits beside the movements, with the actions to resolve it.
- Exceptions are raised automatically when something mismatches or a file goes missing.
- They close themselves when the underlying problem is fixed.
- Assignment, status and discussion live on the exception's own thread.
Why it doesn't balance
Processor
$5,000.00
Bank
$5,031.20
Difference
-$31.20
Matched on
payout_id = bank_reference
po_9f3a2c71
Capabilities
Built around how money actually moves.
One-to-one and one-to-many
Match a payout to its bank deposit, or a settlement total to the thousands of card clearing records behind it. Partial groups wait for the rest to arrive, then get flagged once they've aged past your limit.
Rules on any field
Match on IDs, references and dates from either side, with fixed values, date windows and formatting differences ignored, and see exactly which values each reconciliation was matched on.
File feeds with a test bench
Set up a CSV, BAI2, JSON or fixed-width feed and try it against a sample file before going live.
Built for scale
Card clearing files run to millions of records, and AnselIQ reconciles millions of them in minutes. If a run is interrupted, it picks up where it left off instead of starting over.
Corrections, never edits
A bad file is reversed, not deleted. Write-offs post adjustment entries so the ledger always shows what happened.
Exceptions that close themselves
Mismatches and missing files resolve automatically when the real fix happens. People handle the judgment calls, with assignment and a discussion thread.
AI matching
AI for the matches your rules miss.
Rules handle the clean cases. When the two sides don't line up exactly, like an invoice number the bank reformatted or a reference with an extra prefix, AnselIQ's AI suggests the likely match. Someone on your team accepts or rejects it, and every decision is recorded.
Runs on your servers
The model runs inside your own environment, so transaction data never leaves your network and no outside AI service is called.
Checked, not trusted
Each suggestion comes with a reason and, where it can, the exact rule that makes the two references equal, verified against the records before anyone sees it.
Learns your rules
As your team accepts suggestions, recurring patterns are offered as matching rules. Add one, and those pairs match automatically from then on, with no AI involved.
Open source
Free to run. Help if you want it.
AnselIQ is open source under the Apache License 2.0. Use it, change it and run it in your own environment, with no license fees. If you'd like a hand designing your feeds and matching rules, deploying it, or adapting it to your data, consulting is available.
- 1Clone the repository and build the images.
- 2Install it on Kubernetes with the installation guide.
- 3Want help? Tell us about your setup using the form.