|
BUILD WITH AI |
BUILD WITH AI · PART 4 OF 5
Build an AI-assisted workpaper prep and review workflow
The goal is not to have AI sign off - it is to arrive at senior review with fewer open items and a cleaner tick-and-tie.
Workpaper prep is one of the more time-consuming parts of a close or audit cycle - not because any single step is hard, but because the work is fragmented: pull a trial balance, tie it to a schedule, write a brief narrative, flag anything that moved more than expected, hand it up for review. Each handoff costs time, and most review notes circle back to the same avoidable gaps: missing cross-references, unexplained variances, narratives that describe what happened but not why.
BUILD WITH AI
Workpaper prep and review, step by step
This workflow takes a workpaper from raw data pull to reviewer-ready draft using a general-purpose AI assistant as the connective layer. You supply the verified figures; the assistant handles the drafting, variance commentary, and gap list.
1 |
Export the relevant schedule or trial balance section to a plain text or CSV format and verify the figures against the source ledger before touching the assistant. |
2 |
Open your general-purpose AI assistant and paste the verified data with a brief context line: account name, period, and the prior-period comparator if variance commentary is needed. |
3 |
Paste this prompt to generate the workpaper narrative and initial variance notes: 'You are helping prepare an audit workpaper section. Below is the account activity for [Account Name] for the period ending [Month, Year]. Prior period balance was [Prior Balance]. Current period balance is [Current Balance]. |
4 |
Review every sentence of the returned narrative against the schedule line by line; correct any figure the assistant restated incorrectly or any explanation that does not match the source data. |
5 |
Copy the assistant's list of flagged or unusual items into a separate review-notes column in the workpaper; resolve each one with a tick-and-tie reference or a preparer comment before marking it cleared. |
6 |
Ask the assistant to read the completed draft narrative and list any cross-references or supporting schedules it would expect to see cited but cannot find in the text. |
7 |
Address each item on that gap list or document why it does not apply, then move the workpaper to the senior review queue. |
A general-purpose AI assistant can sit in the middle of that process and do the connective work - drafting the narrative from structured data you paste in, generating a variance comment from two numbers you supply, or reading a draft workpaper section and listing what a reviewer is likely to ask. It does not access your ledger, it does not sign anything, and it cannot verify a figure it did not see. What it can do is compress the gap between a raw data pull and a reviewer-ready document, which is where most of the cycle time actually lives.
The failure mode is straightforward: the AI produces confident-sounding language around a figure you pasted in incorrectly, or it misses a line because you did not include it in the prompt. Neither of those is the AI's fault, but both will show up as review notes. The discipline the workflow requires is the same discipline good workpaper prep has always required - trace every figure back to a source, and do not let a fluent narrative substitute for a verified number.
Thomson Reuters has described a trajectory where AI moves from answering narrow questions toward triggering compliance steps autonomously; for now, the practical value in workpaper work is at the earlier end of that range - drafting and gap-spotting with a human confirming every output against the source data before the workpaper moves forward.
WORKED EXAMPLE
In practice
A staff accountant is preparing the prepaid expenses workpaper for the month just ended. The schedule shows four prepaid lines with current and prior balances; two lines moved by more than 15% and one new line was added this period.
What came back. The assistant produced a two-sentence narrative describing the overall increase and identifying Prepaid Marketing Contracts as a new addition requiring a contract reference and amortization schedule. It flagged Prepaid Insurance as showing a notable increase and noted the software license balance moved in the opposite direction from the other lines, though it described that decrease as likely immaterial without providing a basis for that judgment - something the senior reviewer would likely push back on.
How it was checked. The preparer traced each line balance to the general ledger export, confirmed the insurance and software figures matched, and added a preparer note on the software line referencing the renewal date from the vendor invoice to preempt the reviewer question.
A constructed example. The prompt is usable as written; the figures show the shape of a result, not a measured one.
WHEN TO USE IT
| WHEN NOT TO
|
WHAT TO TAKE FROM THIS
| Paste only figures you have already verified; the assistant cannot catch an input error. | |
| Use the AI's gap list as a checklist, not a clearance - every flagged item needs a human resolution. | |
| Narrative fluency is not accuracy; read every drafted sentence against the underlying schedule. |
SPONSORED
QUESTIONS THIS ANSWERS
Can an AI assistant access my accounting system to pull workpaper data automatically?
A general-purpose assistant cannot connect to your ledger or ERP on its own. You supply the data by pasting in figures or uploading a file; the assistant works only with what you give it.
What part of workpaper prep is AI least reliable for?
Anything that requires tracing a figure to a source document. The assistant drafts and summarizes based on what you paste in, but it cannot verify that the number you pasted matches the underlying record.
Where does AI fit in the review cycle - before or after senior review?
Before. The workflow is designed to reduce the open items a senior reviewer would otherwise catch, not to replace that review.
SOURCES
Where this comes from
The career portal does the rest.
GO DEEPER
Go deeper
IN THIS SERIES
Previously: One prompt, one revision: engagement letters faster
Next: Build with AI: client advisory prep (coming)
An original workflow written for practitioners. Replicate it in a sandbox first; nothing here replaces your review.
How accountants are using AI, automation and smarter workflows to close faster, audit cleaner, and free up time for real work.
Accounting Stack · Audit Friendly Data · Accounting & Finance Jobs
Audit Friendly · modernaccounting.ai