Duplicate payment94%
Same $67.43 Staples charge — receipt and Venmo reimbursement, three minutes apart.
Audit reconciliation for messy nonprofit books
The books are a mess.
The filing deadline isn't.
Ledgerline runs a swarm of specialist agents on your real sources — receipt photos, P2P screenshots, bank CSV, chaotic spreadsheet — and reconciles them with an auditor-grade trail. AI proposes. Code verifies.
4agent roles
4source types
0trusted LLM math
The gap nobody builds for
Budgeting apps categorize.
Ledgerline reconciles.
QuickBooks assumes clean entry. Monarch assumes one linked account. Neither ingests a folder of Venmo screenshots and asks: does this actually match the ledger?
Small nonprofits — youth clubs, community orgs, volunteer-run 501(c)(3)s — keep books across four sources that never talk to each other. At year-end, someone spends nights matching receipts by hand.
“We need reconciliation, not another pie chart.”
bank_statement.csv
receipt_scan.jpg
venmo_feb.png
ledger_v3.xlsx
One trail. Every decision proven.
The forensic catch
It finds what a tired treasurer would miss.
Three concrete anomaly classes — not vibes. Each flag ships with evidence, confidence, and a recommended action your board can act on.
Category mismatch91%
Spotify Premium filed under program supplies. Personal subscription, not a program expense.
Missing documentation88%
$240 ledger entry with no matching receipt or bank line above the documentation threshold.
How it works
Three steps. One reconciled ledger.
Drop the shoebox
Bank CSV, receipt photos, Venmo screenshots, treasurer spreadsheet — one folder, four sources, zero manual entry.
Watch the swarm argue
Ingestion parses. Matcher proposes pairs. Skeptic challenges weak matches. You see every handoff — not a black box.
Export the trail
Anomalies in plain English. Ledger math recomputed in code. One click to a board-ready PDF with full provenance.
Architecture
A swarm that argues — not a chatbot that guesses.
Four role-specialized agents with explicit handoffs. The Matcher proposes; the Skeptic challenges; the Reconciler never trusts AI for arithmetic. Every decision logged with provenance.
Ingestion
OCR + CSV → clean transactions
Matcher
Cross-source pairs + confidence
Skeptic
Challenges matches, runs anomaly taxonomy
Reconciler
Deterministic ledger verification
Deterministic ledger backstop
Live agent trace playback
Board-ready PDF export
Fixed anomaly taxonomy
Your books are already messy enough.
Load our sample nonprofit dataset or upload your own.
No account. No bank linking. Just reconciliation.