Accounts receivable is where finance teams lose the most hours to work that feels necessary but rarely feels productive. Chasing late payers, reconciling partial payments, rebuilding the same aging report every week, and drafting reminder emails that all say the same thing in slightly different words — this is the daily reality for most controllers and AR analysts. The work is repetitive, deadline-driven, and unforgiving of error. A single misclassified invoice can distort the aging buckets, and a missed follow-up can push a 30-day delinquency into a 90-day problem.
The deeper friction is not the volume of tasks — it is the context switching. An AR analyst needs to know a customer’s payment history, their contract terms, the current invoice status, and the tone of previous correspondence before sending a single collection email. Multiply that by hundreds of open invoices and the cognitive load becomes unsustainable. Teams respond by standardizing everything into generic dunning templates, which customers learn to ignore, or by doing nothing until an invoice crosses a threshold that triggers escalation.
ChatGPT does not replace the AR ledger or the ERP. What it does is compress the preparation work: drafting stage-appropriate collection messages, summarizing customer payment behavior, structuring aging commentary for management review, and converting raw exported data into a clean, decision-ready format. Used with a disciplined prompt structure, it turns a 40-minute weekly reporting task into a 10-minute review-and-approve task.
Where the leverage actually sits
The highest-value use cases are not the flashy ones. They are the routine outputs that consume senior attention: the weekly aging narrative for the CFO, the escalation email for a customer who has ignored two reminders, the summary of disputed invoices before a collections call. These are structured, repeatable, and benefit enormously from a consistent prompt template that encodes your tone, your escalation logic, and your reporting standards.
The two prompt templates below are built for that purpose. The first handles collection correspondence across escalation stages. The second handles the aging report commentary and bucket analysis. Both follow the same anatomy: task, success criteria, context files, reference pattern, success brief, and a mandatory clarification step before execution.
First, read these files completely before responding:
[collections_policy.md] — our escalation stages, grace periods, and approved tone at each stage
[customer_master_export.md] — customer names, account tiers, payment terms, and relationship notes
[open_invoices_export.md] — invoice numbers, amounts, due dates, and days past due
Here is a reference for what I want to achieve:
[Upload a previously approved collection email as markdown, or describe the tone and structure of one that worked]
Here’s what makes this reference work:
[Extract: subject line pattern, opening line that names the invoice without accusation, single clear ask, escalation language calibrated to stage, sign-off that preserves the relationship]
Here’s what I need for my version / SUCCESS BRIEF:
Type of output + length: Three email drafts (Stage 1 reminder, Stage 2 firm follow-up, Stage 3 pre-escalation notice), each under 150 words
Recipient’s reaction: The customer should recognize the invoice immediately, understand the consequence of further delay, and feel the message is professional rather than hostile
Does NOT sound like: Generic dunning boilerplate, legal threats at Stage 1, or overly apologetic language at Stage 3
Success means: Our AR analyst can send each draft with fewer than two edits and the escalation logic matches our policy document exactly
My context file contains my standards, constraints, audience. Read it fully before starting.
DO NOT start executing yet. Ask clarifying questions first.
Give me your execution plan (5 steps max) before you begin.
Notice what the first prompt does structurally. It does not ask ChatGPT to “write a collection email.” It defines the escalation stages, supplies the policy and data files, anchors the tone to an approved reference, and specifies a measurable success condition — fewer than two edits per draft. That last constraint is what separates a usable prompt from a demo prompt. It forces the model to internalize your standards rather than invent its own.
The second prompt targets the reporting side. Aging reports are numerically simple but narratively difficult: the CFO wants to know why the 60-plus bucket moved, which customers drove the change, and what action is being taken. Building that commentary manually every week is where controllers lose their Friday afternoons.
First, read these files completely before responding:
[current_aging_export.md] — this week’s aging buckets by customer and invoice
[prior_aging_export.md] — last week’s aging data for movement comparison
[ar_commentary_standards.md] — our reporting format, materiality thresholds, and required sections
Here is a reference for what I want to achieve:
[Upload a prior week’s approved aging commentary as markdown, or describe its structure]
Here’s what makes this reference work:
[Extract: opening summary sentence pattern, how bucket movements are quantified, how top drivers are named with amounts, how recommended actions are phrased as owner-plus-deadline, and the closing risk statement]
Here’s what I need for my version / SUCCESS BRIEF:
Type of output + length: A 300-400 word aging commentary with four sections — headline summary, bucket movement analysis, top five delinquent accounts with drivers, and recommended actions
Recipient’s reaction: The CFO should be able to read it in two minutes and know exactly where the risk sits and who owns the next step
Does NOT sound like: A data dump, a list of every invoice, or commentary that restates the numbers without explaining them
Success means: The commentary passes CFO review with no more than one round of revisions and every recommended action has a named owner and date
My context file contains my standards, constraints, audience. Read it fully before starting.
DO NOT start executing yet. Ask clarifying questions first.
Give me your execution plan (5 steps max) before you begin.
The critical discipline with both prompts is the clarification step. Do not skip it. When ChatGPT asks questions before executing, it surfaces gaps in your own inputs — a missing policy clause, an ambiguous materiality threshold, a customer tier you forgot to define. Those questions are the cheapest quality control you will get all week.
Start with the aging commentary prompt. It has the highest ratio of time saved to risk taken, because the output is internal and reviewed before distribution. Once your team trusts the format, move to the collection sequence and calibrate the escalation language against your actual policy. Keep your context files current — a stale customer master will produce confidently wrong drafts, and that is the one failure mode worth guarding against.
Published on 22 October 2026 on growwithgpt.com