AI for Accountants and Auditors: Uses and Review Controls
How accountants and auditors use AI for preparation, research, and documentation while maintaining human judgment at every decision point, with a record of responsibilities and decisions outside the financial system.
Updated 5 min read
AI supports accountants and auditors in two broad areas: preparing documentation and research materials more quickly, and structuring the questions and tests that professional judgment then applies. The line between these areas matters. Preparation can be accelerated by AI; the professional determination that comes after it cannot be outsourced.
Tools like fenbs keep the decisions about who is responsible for what, the open questions, and the review commitments in a place outside the financial records, so the team works from the same understanding of what has been done and what remains.
Accounting preparation: where AI adds speed
For accountants, the preparation tasks that AI handles well are the writing and structuring tasks: drafting client communications, summarizing accounting standards, generating working paper structures, and writing explanations of complex concepts in plain language for client review.
AI can also help generate reconciliation structures, produce a first-draft checklist for a close process, and write procedure narratives for documentation. The accountant reviews, corrects, and approves each output. The professional standard requires that the work product reflect qualified judgment, not just AI generation.
Numbers must be verified from source. AI produces incorrect figures with the same confidence as correct ones. Any amount in a working paper, a financial statement, or a client report needs a traceable source that a human reviewed. Do not include AI-generated numbers without independent verification.
Audit planning: AI-supported, judgment-led
In audit, AI is useful at the planning stage for tasks like generating a list of inherent risk factors relevant to an industry, structuring an analytical procedure approach for a particular balance, or summarizing relevant auditing standards (such as AS 2110, AS 2301, and related PCAOB or AICPA standards for the engagement type).
The risk assessment and materiality determination that follow are professional judgments. AI can help you think through risk factors systematically; it cannot make the determination that a risk is significant or that a balance is material. Those decisions require the auditor's knowledge of the entity, the industry, and the specific engagement.
| Task | AI-supported (preparation) | Human judgment required (determination) |
|---|---|---|
| Risk identification | Generate a list of inherent risk factors for the industry and account type | Assess which risks are significant for this specific entity and engagement |
| Analytical procedures | Suggest analytical approaches for a balance; structure the comparison | Interpret results; determine whether fluctuations require further investigation |
| Audit program | Draft procedure templates from relevant standards | Select which procedures address the identified risks; approve the final program |
| Documentation | Draft narratives and memo structures for working papers | Review for accuracy; sign off that the work reflects what was actually done |
| Standards research | Summarize relevant PCAOB or GAAS standards | Apply standards to the specific facts; cite the source directly, not the AI summary |
Evidence review: AI cannot substitute for auditor procedures
Auditing standards require the auditor to obtain sufficient appropriate audit evidence. AI tools are not a substitute for that evidence. An AI summary of a contract is not the same as reading the contract. An AI description of a process is not the same as observing or testing it.
Where AI can assist is in processing large volumes of data for anomaly identification, sampling support, and document summarization before the auditor reviews. In all cases, the auditor applies professional skepticism to the AI output and tests key items directly.
Audit firms are developing firm-specific AI tools with controls designed for the audit context. Check your firm's current guidance on approved tools and procedures before incorporating any general-purpose AI tool into audit work.
Record review responsibilities and open questions
fenbs' Decisions and rules section keeps the engagement decisions, the review assignments, and the open questions that need resolution, outside the financial records. When a different team member picks up a task, they read who is responsible for each review step and what questions are still open. People decide; the record captures it.
Create a free accountFree for one person · No credit card · Works with Claude, ChatGPT, Cursor and more
Review controls: what to document
Any AI-assisted work product should have a clear record of what was AI-generated and what was reviewed and approved by a qualified person. This is not a new principle; it reflects the professional responsibility that has always applied to supervised work. The practical implication is that a memo drafted by AI and reviewed by a CPA is the CPA's work product.
Document the review step, not just the output. "Drafted by AI, reviewed against ASC 606 by [name] on [date]" is a useful record. "Generated" with no review record is not. The review is what makes the work product defensible.
Keeping a record outside the financial system
Engagement decisions, who is responsible for which audit area, how a judgment call was resolved, and why a particular approach was chosen, belong in the audit file. But the coordination record, the open questions, the outstanding items, and the status of each task area, often live in emails and conversations that are hard to trace.
A shared project record outside the financial records keeps that coordination visible to every team member and every AI session. It does not contain client financial data; it contains the procedural decisions and status information that help the team work coherently.
How to do this with fenbs
- Create a free account and add the engagement as a project. Include the entity name, the engagement type (audit, review, compilation, tax), and the current stage.
- Add tasks for each major work area. Note who is responsible and what remains open. Do not include financial data, client account numbers, or PII.
- Record the review assignments and engagement-level decisions in Decisions and rules: who signs off on each area, how judgment calls are escalated, and any scope decisions made.
- Use Links and listings for references to standards documents, firm guidance, and client-provided documentation, with the date accessed and the issue they address.
- Connect a supported AI app from Integrations and ask it to read the project context at the start of each session. It starts from the current engagement state, not from a blank page.