How it works

01 Upload

Drop client files into the data room. Bank statements, invoices, trial balances, receipts — any format your client sends.

02 Parse

The coworker reads every document, extracts structured data with page-level citations, and flags anything it can't confidently classify.

03 Generate

From extracted facts: journal entries, reconciliation matches, and workpaper schedules — all linked to source documents.

04 Review

Your team reviews the output. Approve line-by-line, adjust classifications, resolve flags. Nothing leaves without sign-off.

Document to workpaper

Client files in
upload.finalize({
  client: "Acme Sdn Bhd",
  engagement: "FY2026 review",
  files: ["bank.pdf", "invoices.zip"],
  hash: true,
  detectDuplicates: true,
});
Reviewed facts out
workpaper.produce({
  source: "approved_facts",
  output: "lead_schedule",
  citations: true,
  format: "reviewable",
});

Capabilities

Multi-format parsing

Invoices, bank statements, receipts, trial balances, ledger exports — PDF, Excel, images. Extracted to structured facts with page-level provenance.

Journal entry generation

Maps extracted facts to your chart of accounts. Suggests debit/credit entries with classification confidence scores. You approve or adjust.

Bank reconciliation

Matches bank lines to ledger entries automatically. Surfaces unmatched items, timing differences, and discrepancies for your review.

Workpaper assembly

Produces reviewable schedules from approved facts — lead schedules, supporting schedules, and tie-out summaries linked to source.

Review queues

Every generated output sits in a queue until a qualified human approves it. Nothing propagates downstream without explicit sign-off.

Audit trail

Full processing history per document — what was read, what was extracted, what was generated, who approved, when. Complete traceability.

Primitives

Data Room collect

Client document collection with hashing, dedup, and version tracking.

Parser extract

Multi-format extraction with structured output and confidence scoring.

Mapper classify

Account classification against your CoA with GL code suggestions.

Reconciler match

Bank-to-ledger matching with tolerance rules and exception surfacing.

Generator produce

Workpaper and schedule generation from approved, cited facts.

Queue review

Approval workflow — nothing propagates without human sign-off.

Upload a client file set. See the full pipeline in under five minutes.

Create a workspace, connect your AI provider keys, and drop in a bank statement or invoice batch. Your coworker handles the rest — you handle the judgment.