Bank reconciliation · reimagined
Reconciliation,
with an AI twist.
Reconcail matches every bank line to your ledger. Rules go first, machine learning comes next, and AI handles only the genuinely ambiguous cases. Every suggestion is explained and every decision is yours.
Illustrative preview
Ticking and bashing doesn't scale
Matching bank lines to ledger entries by hand is slow, error-prone and hard to audit, and the exceptions turn up late.
References are buried
Payer names and references are hidden in bank narratives that never quite match the ledger.
Money moves in pieces
One deposit covers several entries, fees are netted off, and FX shifts the amounts.
Breaks surface at month-end
Genuine exceptions are found weeks late, after hours of manual checking.
Acquire. Analyse. Act.
One reconciliation journey, from statement to sign-off, that gets more automatic with every review.
Acquire
Bank statements and ledger extracts arrive on their own, via bank feeds, SFTP or browser automation. Any CSV or Excel layout is labelled and normalised automatically.
Analyse
Each bank line goes through rules, then ML, then AI. The result is an explained, confidence-scored match suggestion, never a silent ledger write.
Act
Accept, reject or correct with a reason. Each decision is audited, and each one improves the next run.
The AI twist: only where it counts
Each line is passed to the next tier only when the cheaper one cannot settle it. The first confident match wins.
Deterministic first
Exact amount and date matches, settlement windows, reference patterns and your own exclusion rules settle most of the volume, cheaply and the same way every time.
Machine learning ranks the rest
A trained model scores candidates on amount, timing, counterparty and history, and recovers matches that rigid rules leave behind.
AI for the genuinely ambiguous
Netted fees, split settlements and FX movements go to an AI resolver. It picks from a bounded candidate set or abstains, and code checks the amounts again before anything is suggested.
Confidence bands
Ticked with a full audit trail.
Queued for a person with the evidence attached.
Surfaced early as a genuine break, not at month-end.
Built for messy real-world money
One bank line, one ledger entry
One deposit settling several entries
Split payments that land as one entry
Batches on both sides
International near-amount linking
Small over and under differences
Reversals, bounce-backs and re-issues
Cash in the bank, not yet in the ledger (and the reverse)
Auditable from the first run
Built for regulated finance teams, where every tick has to stand up to an auditor.
AI advises. People decide.
Reconcail never writes to your ledger. Every suggestion is a recommendation that a person accepts, rejects or corrects.
Explainable by default
Every match carries its tier, rule, confidence, features and a plain-English reason, exported as JSON, CSV and a readable report.
Every row accounted for
Integrity checks confirm that every bank and ledger row is either matched, excluded with a reason, or raised as an exception.
Safe AI by design
Bounded candidate sets, amounts checked again in code, prompt-injection resistance and a circuit breaker. If the AI is unavailable, the run finishes on rules and ML.
Proven against your process
Back-test on your own history and measure agreement with your current spreadsheet or macro before you switch.
Runs in your cloud
Serverless and event-driven. Your data stays in your account and region, and reports are written to immutable storage.
Made for
Turn checkers into exception managers.
Bring three months of bank and ledger exports. We'll back-test Reconcail against your current process and show you the results.
Request early access