---
title: "Accounts receivable: 90-day AI follow-up vs bad-debt write-off"
site: "mm-ais.com"
kind: "article"
author: "Claire Dawson"
date: "November 19, 2026"
canonical: "https://mm-ais.com/blog/accounts-receivable-90-day-ai-follow-up-vs-bad-debt-write-off.php"
markdown: "https://mm-ais.com/blog/accounts-receivable-90-day-ai-follow-up-vs-bad-debt-write-off.php/index.md"
description: "Compare 90-day AI follow-up with bad-debt write-off for accounts receivable, and see why static statements fail to adapt to payment, disputes, or silence."
---

# Accounts receivable: 90-day AI follow-up vs bad-debt write-off

Claire Dawson · November 19, 2026

> Compare 90-day AI follow-up with bad-debt write-off for accounts receivable, and see why static statements fail to adapt to payment, disputes, or silence.

Static monthly statements are a weaker default because they summarize the account without changing the next action. As quickreceivable.com explains, an accounts receivable statement lists invoices, payments, credits, and an aging summary, and the three main statement types are open item, balance forward, and activity statements. Those formats help a customer reconcile the account, but none of them adapt to whether the customer paid, disputed, or ignored the last reminder.

Premature write-off loses recoverable cash. A balance with a partial payment, payment plan, or open dispute is not the same as an uncollectible balance; writing it off before the 90-day and three-attempt test removes the follow-up path that might still convert it to cash. The write-off review should be reserved for balances that have aged past 90 days and have no documented customer response.

Use the table as a weekly control: sort the aging report by bucket, then by last response date, and move only the 90-plus-day, zero-response rows into write-off review. Every other row stays in the AI follow-up queue, even when the balance is old enough to look doubtful on a static statement.

![Rain slick city finance district dusk glass towers reflecting](https://static.mm-ais.com/article-images-ai/accounts-receivable-90-day-ai-follow-up-ai-7164bce9.jpg)
Rain slick city finance district dusk glass towers reflecting

## Cost of being wrong in both directions

The cost of being wrong is not symmetric. Over-following spends effort on balances that will not pay; under-following gives up cash that could still be collected and leaves the allowance too low. The two errors pull the same receivable in opposite directions, so the control has to separate collection effort from accounting classification.

Over-following is cheapest at the margin—an automated touch costs little—but it is not free. Once a balance passes 90 days with no response, continued chasing consumes collection-staff time and postpones the allowance adjustment that keeps the financial statements aligned with expected collections. The error is not the touch itself; it is failing to stop, document the silence, and move the balance into review.

Writing off a 61-90 day balance that still has an active dispute turns a recoverable amount into expense before collection is no longer reasonably assured. The direct-matching principle points to recognizing expense when recovery is no longer reasonably assured, not when the balance is merely old. The cost is both the cash that could have been collected and the understated allowance that results from removing a balance too early.

Waiting too long has the opposite effect: receivables stay overstated, the allowance stays understated, and the eventual expense lands in a later period than the related revenue. That timing mismatch can make a weak collection position look stronger than it is until the write-off finally occurs, and it compresses the expense into a period that did not earn the related revenue.

A useful monthly review is to list every balance that has crossed 90 days, note the last customer response, and decide whether it needs one more documented attempt or a write-off review. If the same balance appears in both the follow-up queue and the write-off review, the process is not separating effort from classification.

| Error | Cost | Check |
| --- | --- | --- |
| Over-following | Staff time and delayed allowance adjustment | Stop automated touches after no response and move to review |
| Early write-off | Forfeits recoverable cash and understates allowance | Keep disputed 61-90 day balances in follow-up |
| Late write-off | Overstates receivables and understates allowance | Classify only after 90+ days and three documented attempts with no dispute |

![Cost of being wrong in both directions — Accounts receivable](https://static.mm-ais.com/article-images-ai/accounts-receivable-90-day-ai-follow-up-ai-d3518470.jpg)

## What the evidence does not establish

The evidence base behind the 2026 default is narrower than the default itself. The available sources sources establish what an AR statement contains, how an aging summary is structured, and that receivables sit on the balance sheet as an asset. They do not supply a measured collection-rate lift for AI-driven follow-up versus a static reminder cadence, and they do not identify an optimal day to write off. Those two inputs have to be backtested against your own AR history before anyone treats them as settled numbers.

Start with the lift question. QuickReceivable’s October 23, 2026 deep read on accounts receivable statements describes the statement as a summary of invoices, payments, credits, and the total balance owed, with an aging summary customers act on most often. Oracle’s statements documentation describes statements as print, file, email, or fax reports showing amounts owed, typically prepared for billing purposes. Neither source reports a percentage improvement from automated, AI-conditioned follow-up over a fixed schedule. If a vendor quotes an X% improvement, the check is to replay that claim on your own invoice-level history: same customer set, same aging buckets, same cash application cutoffs, and a measured difference in collection timing rather than a headline rate.

The second gap is the write-off trigger. The aging buckets themselves are sourced — current, 1 to 30, 31 to 60, 61 to 90, and over 90 days past due, as QuickReceivable lays out — but the number of documented touches before classification is not. The three-attempt rule in this playbook is a stated operational default, not a benchmark drawn from the available sources. Treat it as a control you can test: hold the 90+ day and no-dispute conditions fixed, then compare outcomes where the third attempt is documented against outcomes where it is not, using your own recovery results as the deciding evidence.

Allowance method selection is a third area the available sources leaves open. The sources cover statement structure and aging presentation; they do not compare the percentage-of-receivables approach with an aging-schedule approach, and they do not tell you which one fits your portfolio. That choice stays with your accounting policy and your own loss history. What the sources do support is the input quality: QuickReceivable notes statements should be generated only after cash has been applied and credit memos posted, which is the same discipline an internal backtest needs before any balance is counted as recoverable or unrecoverable.

The explicit limits are summarized here, once:

| Question | What the available sources establishes | What you must supply internally |
| --- | --- | --- |
| Collection lift from AI follow-up | Statement contents, aging summary, statement delivery formats | Backtested lift on your own AR history; no sourced percentage exists |
| Optimal write-off day | Aging bucket structure, including over 90 days past due | Backtested recovery timing; no sourced optimal day exists |
| Touches before classification | Documentation of invoices, payments, credits, and disputes on the statement | Your three-attempt default, validated or revised from your own results |
| Allowance method | Aging presentation and statement preparation sequence | Method selection from your policy and loss history; not covered by the available sources |

The practical consequence is that the default stays a default until your data says otherwise. Run the follow-up sequence as written, keep the attempt log and dispute status on every balance, and let the backtest decide whether the lift and the timing assumption hold in your portfolio.

## $48,000 aging to decision

Here is the single worked example for the method, using one customer with a $48,000 total balance. The starting aging report shows $12,000 current, $10,000 in 1–30, $8,000 in 31–60, $6,000 in 61–90, and $12,000 over 90. Those figures are illustrative inputs for the decision method, not sourced data. The check is arithmetic: $12,000 + $10,000 + $8,000 + $6,000 + $12,000 = $48,000.

At day 30, the AI-driven sequence has sent the statement and logged no response on the $24,000 that is now past due. That $24,000 is the sum of the original 1–30, 31–60, and 61–90 buckets: $10,000 + $8,000 + $6,000 = $24,000. The $12,000 already over 90 is separate. The operational check is not the tone of the message; it is whether the statement, payment, and dispute fields reconcile to the customer account before the next follow-up step is released.

At day 60, the escalation notice goes out and a $4,000 partial payment is received. Apply the cash before judging the balance. The total customer balance becomes $48,000 − $4,000 = $44,000. The decision rule at this checkpoint is to keep the AI follow-up running on every remaining invoice through day 90, because the partial payment is evidence of engagement and the remaining balance has not yet met the write-off rule.

At day 90, three documented follow-up attempts are complete, and the $12,000 over-90 balance remains unresponsive with no dispute. That is the decision point in this example. The journal-entry decision is to write off the $12,000 over-90 balance against the allowance for doubtful accounts, recording the reduction in the receivable and the corresponding use of the allowance rather than treating the entire $44,000 remaining customer balance as uncollectible.

The final check is that the buckets still tell a coherent story after the entry. Before the write-off, the remaining balance is $44,000; after removing the $12,000 over-90 balance, the book balance is $32,000. The $4,000 partial payment has already reduced cash timing pressure, and the write-off is limited to the balance that reached the documented day-90 condition. This is why the aging-to-decision workflow matters: it separates balances still being actively worked from the specific amount that qualifies for the allowance entry.

## Worked Example: Run the Numbers

Here is the single worked example this section carries end to end. Illustration only: Northline Supply issued an invoice to Vauxall Holdings on March 1, 2026, for a delivered order, and the balance remained open when the October 23, 2026 customer statement described by QuickReceivable was generated. The statement lists the invoice, any payments and credit memos, the closing balance, and the aging summary — it does not create a new amount owed, per that same source. We are testing one question: does the 90-day AI follow-up cadence move cash timing on this balance, or does the balance sit until a write-off review is justified?

Illustration only — step one, age the balance. From March 1 to October 23 is 236 days. That places the invoice beyond the 61 to 90 band and into the over-90 bucket of the aging summary QuickReceivable documents (current, 1 to 30, 31 to 60, 61 to 90, over 90). The check here is mechanical: pull the customer statement only after cash has been applied and credit memos posted, because QuickReceivable notes statements should be generated at that point. If a payment or credit memo is sitting unapplied, the aging bucket is wrong and every downstream decision is wrong with it.

Illustration only — step two, apply the reader rule. Because this balance is over 90 and the rule requires at least three documented follow-up attempts with no dispute before bad-debt classification, we count the log. Suppose the AI sequence recorded three attempts and the customer raised no dispute on any of them. Three documented attempts, zero disputes, over-90 status: the classification condition is met. If the log showed two attempts, or a dispute on the third, the balance would stay in follow-up rather than move to classification — the count and the dispute flag are both required, not either one.

Illustration only — step three, the timing comparison. The AI-driven sequence ran across the full 90-day window, so any cash it pulled forward arrived before the classification date. On this fact pattern the winner is the follow-up sequence, not the write-off: it produced cash timing movement on a balance that was still recoverable during the run. The break-even trigger is the third documented attempt with no dispute — cross that line and the balance stops being a follow-up candidate and becomes a classification candidate. Stay short of it and the same balance stays in the cadence.

The arithmetic to verify before you close the file: 236 days from March 1 to October 23, 2026; one invoice; three attempts; zero disputes; one bucket (over 90). Recompute the day count against your own statement date rather than copying this one — the bucket assignment depends on it. And keep the statement itself in the file, since QuickReceivable's format includes the aging summary and a contact for billing questions or disputes, which is exactly the evidence the three-attempt/no-dispute test needs.

## Decision rules for the 90-day trigger

The decision rules work as a single if/then sequence, and the first branch is the one most teams get wrong. If a balance is under 90 days past due, keep it in the AI follow-up lane: send the monthly statement described in QuickReceivable’s October 23, 2026 guidance, then let any customer response trigger escalation to a collector. No income-statement entry belongs in this branch. The statement itself does not create a new amount owed; it summarizes invoices, payments, credits, and the closing balance, so the work is about changing cash timing rather than recognizing expense.

Two exceptions override the straight aging path, and both should be coded into the AI workflow before the first statement goes out. If the customer disputes any portion of the balance at any bucket, reset the aging clock for the disputed amount and route that amount to collections. If the customer makes a partial payment at any bucket, treat the unpaid remainder as the live balance, reset the aging clock for the amount still contested or unresolved, and route it to collections as well. The point is not to reward delay; it is to stop the system from treating a live conversation as silence.

Documentation is the control that makes the write-off defensible. Each attempt should show the date, the statement or message sent, the channel, and the response status, including “no response” where applicable. Before releasing a statement or advancing an account, confirm that cash application is complete and credit memos are posted; the available source describes statement contents, so treat this posting check as an internal control. A partial payment or open dispute keeps the balance in follow-up rather than write-off review.

 credit that has not been posted can make an aging position look older than it really is. Reconcile cash application before the third attempt is counted.

Operationally, the ruleset is: under 90 days, run AI follow-up with monthly statements and response-triggered escalation; 90+ days with three documented attempts and zero response, write off against the allowance and recognize bad-debt expense only through the allowance estimate; dispute or partial payment at any bucket, reset the aging clock for the affected amount and route it to collections. That is the whole trigger logic. It keeps recoverable balances in the cash-timing workflow and reserves expense recognition for the balances that have actually failed the documented follow-up test.

## What to do next

| Step | Action | Why it matters |
| --- | --- | --- |
| 1 | Open the AR aging and flag every invoice that has not reached 90 days past due for the 2026 AI follow-up queue. | AI follow-up is the 2026 default through 90 days and affects cash timing on the balance sheet. |
| 2 | Run the AI follow-up sequence on each invoice through 90 days, logging every touch against the customer account. | The canonical rule requires follow-up on every invoice through 90 days before any bad-debt classification. |
| 3 | At 90+ days past due, check the account for at least three documented follow-up attempts and confirm there is no dispute, promise-to-pay, or partial payment. | The 90+ day bucket is not enough by itself; failed documented attempts with no dispute are required. |
| 4 | Keep any balance that still has a dispute, promise-to-pay, or partial payment in accounts receivable rather than moving it to bad-debt expense. | Accounts receivable is a balance sheet asset, not an income statement line. |
| 5 | Only after the 90+ day and three-attempt tests are met, write the balance off against the allowance, not revenue. | Bad-debt expense hits the income statement only after you write off the receivable. |
| 6 | Reconcile the write-off entry to the allowance and confirm no remaining open invoice in the 2026 queue was expensed before 90 days. | This preserves the decision rule and keeps expense timing tied to write-off, not aging alone. |

## Frequently Asked Questions

**Why shouldn't a monthly statement be the default follow-up tool?**

Static monthly statements are a weaker default because they summarize the account without changing the next action.

**What do open item, balance forward, and activity statements actually show?**

An accounts receivable statement lists invoices, payments, credits, and an aging summary, and the three main statement types are open item, balance forward, and activity statements.

**What can't statement formats tell you about the last reminder?**

Those formats help a customer reconcile the account, but none of them adapt to whether the customer paid, disputed, or ignored the last reminder.

**Which balances should not be written off just because they look old?**

A balance with a partial payment, payment plan, or open dispute is not the same as an uncollectible balance; writing it off before the 90-day and three-attempt test removes the follow-up path that might still convert it to cash.

**How should the aging report be sorted for the weekly write-off control?**

Use the table as a weekly control: sort the aging report by bucket, then by last response date, and move only the 90-plus-day, zero-response rows into write-off review.

**What happens to rows that don't qualify for write-off review?**

Every other row stays in the AI follow-up queue, even when the balance is old enough to look doubtful on a static statement.

## Quick answers

| Why are static monthly statements considered a weaker default for accounts receivable? | They are a weaker default because they summarize the account without changing the next action and do not adapt to whether the customer paid, disputed, or ignored the last reminder. |
| --- | --- |
| When should a balance be reserved for write-off review? | A balance should be reserved for write-off review only when it has aged past the specified period and has no documented customer response. |
| Why should balances with partial payments, payment plans, or open disputes not be written off prematurely? | These balances are not the same as uncollectible balances, and writing them off before the required follow-up period and three-attempt test removes the follow-up path that might still convert them to cash. |
| How should the aging report be used as a weekly control? | Sort the aging report by bucket, then by last response date, and move only the rows past the specified aging period with zero response into write-off review while keeping every other row in the AI follow-up queue. |
| What is the difference in cost between over-following and under-following? | Over-following spends effort on balances that will not pay, while under-following gives up cash that could still be collected and leaves the allowance too low. |

Also worth reading: **Accounts Payable and Receivable Explained Simply**: [Accounts Payable and Receivable Explained](https://mm-ais.com/blog/accounts-payable-and-receivable-explained-simply.php) · **7 Time-Saving Excel Formulas to Automate Your Monthly Expense Reports in 2024**: [7 Time-Saving Excel Formulas to](https://mm-ais.com/blog/7_time_saving_excel_formulas_to_automate_your_monthly_expens.php) · **Streamline Your Business Travel and Expense Management**: [Streamline Your Business Travel and](https://mm-ais.com/blog/streamline-your-business-travel-and-expense-management.php)

### Related reading

- [Accounts Payable and Receivable Explained Simply](https://mm-ais.com/blog/accounts-payable-and-receivable-explained-simply.php)
- [Mastering the Essential Balance Sheet Accounts for Financial Clarity](https://mm-ais.com/blog/mastering-the-essential-balance-sheet-accounts-for-financial-clarity.php)
- [The Top Free Options For Setting Up Business Email Accounts](https://mm-ais.com/blog/the-top-free-options-for-setting-up-business-email-accounts.php)
- [Step-by-Step How to Create a Gmail Alias Without Using Extra Accounts in 2024](https://mm-ais.com/blog/step_by_step_how_to_create_a_gmail_alias_without_using_extra.php)
- [Retail reorder thresholds: fixed (s, Q) default, RL only after backtest](https://mm-ais.com/blog/retail-reorder-thresholds-fixed-s-q-default-rl-only-after-backtest.php)
- [Adaptive vs fixed email cadence: test reply lift against opt-out break-even](https://mm-ais.com/blog/adaptive-vs-fixed-email-cadence-test-reply-lift-against-opt-out-break-even.php)

### Latest

- [Retail reorder thresholds: fixed (s, Q) default, RL only after backtest](https://mm-ais.com/blog/retail-reorder-thresholds-fixed-s-q-default-rl-only-after-backtest.php)
- [Adaptive vs fixed email cadence: test reply lift against opt-out break-even](https://mm-ais.com/blog/adaptive-vs-fixed-email-cadence-test-reply-lift-against-opt-out-break-even.php)
- [Outlook out of office replies: 24-Hour Window vs Artificial Intelligence Sales...](https://mm-ais.com/blog/outlook-out-of-office-replies-24-hour-window-vs-artificial-intelligence-sales-follow-ups.php)

Canonical: https://mm-ais.com/blog/accounts-receivable-90-day-ai-follow-up-vs-bad-debt-write-off.php
Markdown: https://mm-ais.com/blog/accounts-receivable-90-day-ai-follow-up-vs-bad-debt-write-off.php/index.md

```json
{"@context":"https://schema.org","@type":"BlogPosting","headline":"Accounts receivable: 90-day AI follow-up vs bad-debt write-off","url":"https://mm-ais.com/blog/accounts-receivable-90-day-ai-follow-up-vs-bad-debt-write-off.php","description":"Compare 90-day AI follow-up with bad-debt write-off for accounts receivable, and see why static statements fail to adapt to payment, disputes, or silence.","datePublished":"November 19, 2026","author":{"@type":"Person","name":"Claire Dawson"},"articleBody":"Static monthly statements are a weaker default because they summarize the account without changing the next action. As quickreceivable.com explains, an accounts receivable statement lists invoices, payments, credits, and an aging summary, and the three main statement types are open item, balance forward, and activity statements. Those formats help a customer reconcile the account, but none of them adapt to whether the customer paid, disputed, or ignored the last reminder. ..."}
```
