Statement imports and AI extraction: turning bank evidence into reliable data

A bank statement may arrive as a spreadsheet or PDF. Importing it can reduce manual transcription, but extraction is only the first step. The imported data still needs to represent the correct account, dates, amounts, signs, and references.

The purpose of import is to make external evidence usable for reconciliation. It does not establish that every extracted line is correct or that it matches an ERP transaction.

Structured files and documents

A spreadsheet usually presents data in columns, but column names, date formats, and debit-credit conventions can differ between banks. A PDF may contain repeated headers, running balances, several pages, or descriptions split across lines.

DNA supports Excel and PDF statement import in bank reconciliation. Its configured AI support can assist with reading bank-statement PDFs. The extraction result should be understood as a representation of the source document, with accuracy that still needs review.

A practical extraction issue

A statement contains a USD 1,250 debit and a USD 12.50 fee on adjacent lines. If extraction combines or misreads them, the imported total may be wrong even though the text looks plausible.

Another common ambiguity is sign convention. A bank's debit description may not correspond to the way an ERP export labels increases and decreases. The actual account movement and amount need to be interpreted consistently.

Evidence that supports confidence

The account identity, statement period, opening and closing balances, and transaction totals provide useful checks. They help determine whether pages or lines are missing and whether the imported direction is coherent.

Agreement of totals is valuable but not complete proof. Two individual errors can offset each other. Important references and unusual amounts still deserve attention before relying on the data for matching.

AI and responsibility

AI can accelerate interpretation of an unfamiliar layout. It should not be treated as an independent source of bank activity. The statement remains the evidence, and the company's review establishes whether the extracted representation is acceptable.

DNA's company AI configuration determines availability of the documented AI-assisted reading. A configuration issue should not be confused with a statement containing no transactions.

From extraction to reconciliation

Once the imported lines reliably represent the statement, reconciliation can determine which ERP records correspond to them. Extraction and matching are separate tasks: a correctly extracted bank line can still be matched to the wrong internal record.

The useful outcome is a faithful, reviewable set of external lines that supports a clear explanation of the bank account.