
Here's the lowdown
Berlin-based dnl launched a U.S. GAAP Disclosure Checklist that reads annual reports, checks required disclosures and links each proposed answer to its source.
The software checks totals, notes, cross-references and prior-year comparisons, then carries completed work into revised client drafts and shows what changed.
dnl says auditors must still evaluate AI output before using it as audit evidence, and the product records who accepted, adjusted and reviewed each answer.
A checklist can become a second job
A disclosure checklist is supposed to catch what a financial statement forgot. The checklist itself can become a second job.
An auditor may need to read a long report, answer hundreds of questions, trace the numbers back to notes and do it again when the client sends a new draft.
German audit technology company dnl thinks an AI agent can handle much of that first pass. The company has entered the U.S. market with a U.S. GAAP Disclosure Checklist for auditors and financial report preparers.
dnl says its platform is already used by 15,000 auditors at 140 firms, including members of the Big Four. The U.S. launch puts an American GAAP checklist on top of that existing system.
The agent reads before it answers
The workflow starts with a company's annual report.
The agent reads the document, checks each required disclosure and proposes answers where it has enough support. If it is not confident, it sends the item to the auditor as a proposal instead of closing the question on its own.
Each answer links back to the page or passage the agent used. That matters because a green checkmark is not useful if the reviewer cannot see why the software gave it.
The product can also compare a financial statement with a firm's own checklist or reporting rules. That gives firms a way to apply internal requirements along with the standard disclosure questions.
It checks the math too
Disclosure work is not only about missing words.
The software checks totals, cross-references, note amounts and prior-year comparisons across PDF files, spreadsheets and report tables. It is designed to flag figures that do not agree while leaving matching numbers alone.
When a client sends a new draft, dnl carries completed work forward. It also shows what changed between the two versions and whether the change affects the review.
That could remove a common frustration. A small wording change should not force the audit team to rebuild the entire checklist, but a changed figure or note reference still needs attention.
The product also has a report-review step for preparers. It can check a finished report against the financial statements and the firm's reporting standards before the document reaches internal review. Comments, number problems and checklist gaps land in one list that can be discussed with the client and exported together.
The auditor still owns the judgment
dnl is careful about where the software stops.
The company's U.S. checklist page says an auditor may use AI-generated answers as audit evidence only after evaluating their relevance and reliability. It points to AU-C 500, as amended by SAS 142, for that evidence judgment.
The same page says the use of AI should be documented in the audit file under AU-C 230. The product records who accepted or changed an answer, who reviewed it and when those steps happened.
That trail can help a reviewer understand the work. It does not make the AI the auditor. A person still has to decide whether the source is complete, whether the answer fits the facts and whether an exception changes the audit conclusion.
The first U.S. coverage is broad
dnl says the checklist currently supports U.S. GAAP reporting for nonpublic companies, from a single company to a private-equity-backed group.
It also covers not-for-profit reporting under Accounting Standards Codification Topic 958 and financial reporting for cities, counties and other local governments.
The official launch release says the broader checklist can cover private and public companies, not-for-profit organizations, state and local governments and specialized industries. Firms will need to confirm that the exact entity type and reporting framework they use are available in the product version they test.
The controls deserve their own checklist
dnl says it has held ISO/IEC 27001:2022 certification since 2024 and undergoes annual surveillance audits. It also says U.S. client data is hosted in the United States, data is encrypted at rest with AES-256 and data in transit uses TLS 1.2 or higher.
Those are useful starting points, but an audit firm still needs its own technology review. Access rights, retention, model behavior, data isolation and the handling of confidential client files should be tested before the tool enters a live engagement.
The product's appeal is easy to see. It can read a draft, check the math, point to evidence and remember completed work when the next version arrives.
The catch is just as important: faster checking does not transfer responsibility. The auditor still has to understand the answer, inspect the evidence and sign the file.
For more plain-English audit and accounting technology news, follow Ledger Lowdown.

