Win back audit time without giving up judgment
Start with repetitive preparation work that a reviewer can verify. Keep evidence evaluation, exception resolution, and approval with the engagement team. This guide gives audit and accounting leaders a framework for choosing and evaluating a first AI workflow.
By Clark · August 21, 2026 · Updated October 14, 2026 · 4 min read
Automate friction before judgment
The best first use cases are repetitive, evidence-heavy, and easy for a reviewer to verify. PBC intake, extraction, matching, workpaper setup, and first-draft preparation can absorb large amounts of staff time while reviewers still need time to investigate exceptions and apply professional judgment.
A narrow workflow also gives the firm a clean way to measure quality. Teams can compare the AI-assisted result with the existing process, identify exceptions, and decide whether the time saved survives review.
- Evidence collection and document organization
- Extraction and matching across repeatable document sets
- First drafts of workpapers, reconciliations, and routine narratives
- Exception identification for reviewer attention
Human review is part of the system
AI output can look complete before it is supportable. Firms need a review design that makes the source, reasoning, changes, and uncertainty visible to the person accepting the work.
That boundary should be explicit. The system can propose, organize, and draft. The engagement team remains responsible for evaluating the evidence, resolving exceptions, applying professional skepticism, and approving the conclusion.
- Keep every material figure connected to source evidence
- Route exceptions and uncertainty to the right reviewer
- Record adjustments, approvals, and final sign-off
- Prevent client data from moving into unapproved tools
A chatbot is not an engagement workflow
General assistants are useful for research, drafting, and analysis. They do not automatically supply the methodology, permissions, evidence model, review path, or retention controls an audit and accounting firm needs around engagement work.
Before expanding usage, decide where the work will live. A controlled audit or accounting workflow should make the firm’s standards repeatable and leave a record another reviewer can inspect without reconstructing the process from prompts and messages.
Prove one workflow, then expand
Choose one recurring workflow with known inputs, a clear reviewer, and enough volume to measure. Define the required evidence and approval path before the pilot begins.
Measure preparation time, review time, exception rates, rework, and whether the final record is easier to inspect. If the pilot only makes drafting faster while increasing review effort, it has not improved the engagement.
Separate a useful framework from an audit requirement
For engagements under PCAOB standards, AS 1215 addresses documenting procedures, evidence, conclusions, and the people who performed and reviewed the work. Other engagements may follow different standards. The sources below provide context; this article does not replace your firm's methodology or the applicable requirements.
NIST's Generative AI Profile offers a voluntary risk-management reference. Clark's pilot suggestions are an editorial evaluation framework, not a claim that a particular tool meets an auditing standard.
Sources and scope
- PCAOB AS 1215: Audit Documentation
Paragraphs .06 and .06A address procedures, evidence, conclusions, and who performed and reviewed the work. This standard applies to PCAOB engagements; it is not a universal rule for every audit. Check the applicable standard and effective date for your engagement.
- NIST Generative AI Risk Management Profile
A voluntary, cross-sector resource for identifying and managing generative AI risks. It is not an auditing standard or a certification of Clark.