Automating Accruals and Cut-Off Checks with AI

Accruals are the part of the close where small errors compound into material misstatements over time. Cut-off — making sure transactions land in the correct period — is the second area auditors examine carefully. Both are repetitive, judgement-light tasks for most line items, which makes them perfect candidates for AI assistance.

This lesson covers the four accrual workflows that benefit most from AI, plus a cut-off review prompt that catches the issues your team would otherwise find a month late.

What You'll Learn

• The four accrual types where AI delivers the most value
• A reusable prompt for accrual estimation memos
• A cut-off review prompt for the last and first five business days of the period
• The exact line between "AI estimated" and "AI assisted" — and why it matters

The Four Accrual Types Where AI Helps

Type 1 — Recurring service accruals. Utilities, telephone, cloud hosting, professional fees on retainer. These accruals follow a predictable monthly pattern with seasonality. AI is excellent at pattern-fitting from 12 to 24 months of history.

Type 2 — Bonus and commission accruals. Tied to formulas based on revenue or margin achievement. The math is mechanical once the rule is documented. AI helps you write the memo explaining the calculation, not perform it.

Type 3 — Unbilled revenue accruals. Service delivered but not yet invoiced. AI helps you draft the supporting narrative and identify outliers in the unbilled aging.

Type 4 — Vendor invoice accruals (GR-IR style). Goods or services received but invoice not yet entered. AI helps draft the accrual memo and review the GR-IR aging for unusual items.

Across these four, AI is not the calculator. It is the writer, reviewer, and pattern-spotter. Your finance systems calculate. AI explains and documents.

Workflow 1: Drafting the Recurring Accrual Memo

For your recurring monthly accruals, drafting a memo for each one is the most common time tax. Here is the prompt that converts a 5-minute manual write-up into a 30-second AI draft.

"Act as a senior accountant. Draft a 100-word accrual memo for [account: e.g., AWS cloud hosting accrual]. Last 12 months actuals were: [paste month-by-month figures]. Current period invoice not yet received. Estimated accrual: $[X], based on [methodology: 3-month rolling average / prior month + growth / contractual minimum]. Tone: factual, audit-defensible. Include: account, period, basis of estimation, comparison to prior accrual, and reversal expected in the following period."

Output is a clean memo that goes straight onto the accrual workpaper. You spend your time on the judgement piece — is the estimate reasonable? — not on writing words.

Workflow 2: Bonus and Commission Accruals

These have an extra step. AI is not your calculator, but it is your translator and reviewer.

Step 1: Calculate the accrual in your spreadsheet or comp system.

Step 2: Ask AI to translate the calculation into a written explanation:

"Below is a bonus accrual calculation. Convert it into a 150-word memo explaining the methodology, the achievement against target, and the resulting accrual. Tone: clear enough that an auditor or new CFO can follow without prior context. Bonus pool: $[X]. Threshold for full payout: Revenue $[A] / EBITDA $[B]. Year-to-date actual: Revenue $[C] / EBITDA $[D]. Achievement percentage: [E]%. Resulting accrual: $[F]. Prior month accrual: $[G]. Movement: $[H]."

Step 3: AI returns a clean memo. You verify against the calculation and sign off.

This pattern — AI explains, finance signs off — applies to almost every commission and bonus accrual you will ever write.

Workflow 3: Unbilled Revenue Aging Review

Unbilled revenue is where revenue recognition meets billing operations, and where things go sideways. AI is excellent at reviewing the aging and surfacing items that need attention.

Prompt:

"Below is my unbilled revenue aging by customer. Columns: Customer, Service Date, Amount, Days Unbilled. Identify: (1) any items more than 60 days unbilled that suggest billing operations are stuck, (2) any customers with multiple unbilled items totaling more than $10,000, (3) the total aged unbilled balance broken into 0-30, 31-60, 61-90, 90+ buckets, (4) a short narrative I can include in the close memo on the state of unbilled revenue."

Use the output as a checklist for what to escalate to your billing team or revenue ops counterpart. The narrative goes straight into your close documentation.

Workflow 4: GR-IR / Vendor Accrual Review

This is the largest accrual for most product companies. AI helps you review the GR-IR aging the same way it helps review unbilled revenue.

Prompt:

"Below is my goods-received-not-invoiced (GR-IR) aging at month end. Columns: Vendor, PO Date, Goods Receipt Date, Amount. Identify: (1) entries more than 90 days old that may need writing back, (2) vendors with unusually high GR-IR balances relative to monthly spend, (3) any GR-IR entries that look like they should have invoiced by now based on the goods receipt date. Then draft a 120-word memo summarising the GR-IR position at period end, the basis for the closing accrual, and any items recommended for write-back."

Combine this with a manual review by your AP team and you have a GR-IR review that takes a third of the time it used to.

The Cut-Off Review

The cut-off review checks that revenue, expense, and cash transactions land in the correct accounting period. This is what your auditors will scrutinise. AI can do a meaningful first pass.

Master cut-off prompt:

"Act as an audit senior. Below is a list of transactions posted in the last 5 business days of [month] and the first 5 business days of [next month]. Columns: Transaction Date, Posting Date, Description, Amount, Counterparty. Flag transactions where: (1) the posting date is in [month] but the underlying activity (service date or goods receipt) appears to be in [next month] based on the description, (2) the posting date is in [next month] but the underlying activity appears to be in [month] (a missed cut-off), (3) the description is too vague to determine the period and requires clarification. For each flag, return: Transaction ID, side of cut-off, severity, and the clarifying question to ask the preparer."

This is one of the highest-value prompts in this course. Most teams find at least one real cut-off error in the first month they run it. By month three the prompt is tuned tightly to your business.

"AI Estimated" vs "AI Assisted"

Here is the distinction your auditors will care about.

AI estimated. AI calculated the accrual amount. This is rare and should be avoided. AI is not a deterministic calculator.

AI assisted. AI drafted the memo, flagged items for review, summarised aging, or wrote explanatory narrative. This is what you are doing throughout this lesson. It is defensible and increasingly standard practice.

When your audit firm asks (and by 2027 they will all ask), the truthful answer is "AI assisted". You retain the calculation in spreadsheets or systems. AI helps you document and review. That is the safe and supportable model.

Save These as a Single Reusable Workbook

Build one document — call it "Month-end accruals and cut-off prompts" — that holds all five prompts above. Open it on close day, paste each one with the data of the day, and you have just shaved 90 minutes out of the close.

Key Takeaways

• AI drafts accrual memos, reviews aging, and writes narrative — your finance systems do the math
• Use specific 12-month historical context in your prompts for the best draft estimates
• The cut-off prompt finds at least one real cut-off error in the first month, every time
• Always describe AI use as "AI assisted" — never delegate the calculation
• Save your five prompts as a reusable workbook for next month's close