The accounts receivable (AR) function is where cash flow lives or dies, yet it remains one of the most manually intensive processes in finance. Your team spends hours each week exporting aging reports, segmenting delinquent accounts, drafting polite-but-firm collection emails, and reconciling payment promises against actual cash received. The friction is real: sales teams promise terms, customers delay payments, and your analysts are buried in spreadsheet gymnastics instead of strategic cash forecasting. The result? Days Sales Outstanding (DSO) creeps up, bad debt reserves grow, and the CFO’s quarterly cash flow projection misses the mark by millions.
ChatGPT changes this equation. By acting as a tireless analytical assistant, it can ingest your aging report, segment customers by delinquency risk, draft personalized collection sequences, and even generate board-ready summaries of AR health—all in minutes. This isn’t about replacing your team; it’s about removing the repetitive, low-judgment work so your controllers and analysts can focus on high-value disputes, relationship management, and cash forecasting. The key is learning how to prompt it with the right structure, context, and constraints—which is exactly what this guide provides.
Below, you’ll find two production-ready prompt templates. The first automates the entire collection email workflow from a raw aging export. The second generates a comprehensive AR aging report narrative with actionable insights. Both are designed for copy-paste use, with placeholders you can adapt to your own ERP data and customer relationships.
Why Your Current AR Process Is Leaking Cash
Before we dive into the prompts, consider the typical weekly cycle. Your AR analyst pulls a CSV from NetSuite or SAP, sorts by invoice date, and manually identifies accounts over 60 days. Then they check CRM notes for prior promises, guess at the right tone for each email, and hope they haven’t missed a key contact. Meanwhile, the controller wants a narrative of why DSO increased by three days, and the CFO wants a risk heatmap for the board. That’s four hours of grunt work before any strategic conversation happens. ChatGPT compresses that to twenty minutes—if you give it the right scaffolding.
The prompts below follow a structured “anatomy” that forces the AI to behave like a senior AR analyst, not a generic chatbot. They include explicit success criteria, file context, reference patterns, and a mandatory planning phase before execution. This prevents the vague, generic output that most people get when they ask ChatGPT to “write a collection email.” Instead, you get tailored, risk-calibrated communications and analytical reports that read like they came from a seasoned finance professional.
First, read these files completely before responding:
aging_report.csv — the raw aging export with columns: customer_id, customer_name, invoice_number, invoice_date, due_date, days_past_due, amount_open, currency, sales_rep, last_payment_date
crm_notes.md — a markdown file with customer relationship context, including prior promises, dispute status, and communication preferences
credit_terms.md — our standard payment terms (Net 30, Net 60, prepaid) and any exceptions for specific customers
Here is a reference for what I want to achieve:
[Upload a sample of 3 previous successful collection emails from your team as markdown, or describe them in detail — e.g., “Our best emails are under 150 words, start with a specific invoice reference, mention the original promise date, and always include a clear call-to-action with a payment link.”]
Here’s what makes this reference work:
Tone is professional but firm; never threatening legal action on first contact
Each email is personalized with the exact invoice number and amount
The subject line always includes the customer name and days past due
The call-to-action is a single, unambiguous request (e.g., “Please remit $X by Friday”)
We always reference the last interaction date from CRM notes to show we’re tracking
Here’s what I need for my version / SUCCESS BRIEF:
Type of output + length: A separate email for each customer over 30 days past due, each 100-150 words, plus a summary table of all customers ranked by risk (high, medium, low)
Recipient’s reaction: They should feel that we are organized and watching closely, but not aggressive or robotic — they should be motivated to pay without escalating to a dispute
Does NOT sound like: Generic “This is a reminder” templates, legal threats, or overly apologetic language
Success means: At least 40% of contacted customers remit payment within 5 business days, and zero customers complain about tone or harassment
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 prompt above works because it forces ChatGPT to triage your portfolio before drafting a single word. It will ask you clarifying questions—like whether to include sales reps on the emails, or how to handle accounts already in dispute—before it produces anything. That planning phase is critical; it prevents the AI from sending a “pay now” email to a customer who has a legitimate billing dispute that your team is already resolving. When you run this, expect a back-and-forth of 2-3 clarifying questions. Answer them honestly, and the output will be remarkably close to what your best AR specialist would write—minus the coffee breaks.
From Raw Data to Board-Ready Narrative: The Aging Report Generator
The second prompt addresses the other chronic pain point: the monthly aging report narrative. Most controllers spend half a day turning a pivot table into a coherent story for the CFO. Why did 60+ day balances rise? Which customers are at risk of write-off? What’s the collection forecast for next month? ChatGPT can produce that narrative—including a risk heatmap, trend analysis, and recommended reserves—if you feed it historical data and your accounting policies. The prompt below structures that process, forcing the AI to act as a financial analyst, not a writer.
Note the difference in this prompt’s success criteria. It asks for a narrative that a CFO can read in under three minutes and immediately understand the key drivers of AR deterioration. It also demands a specific output format—a summary, a risk table, and a recommended action plan—which mirrors what your board pack requires. This is where ChatGPT shines: not in replacing your judgment, but in compressing the time from raw data to decision-ready insight.
First, read these files completely before responding:
aging_summary_current.csv — the current month’s aging buckets (current, 1-30, 31-60, 61-90, 90+) with dollar amounts and customer counts
aging_summary_prior.csv — the prior month’s aging buckets for comparison
collections_history.csv — actual cash collected per week for the last 12 weeks
write_off_policy.md — our policy for bad debt reserves and write-off thresholds (e.g., automatic reserve at 120 days, write-off at 180 days)
customer_segments.md — customer segmentation by industry, size, and strategic importance (key accounts vs. transactional)
Here is a reference for what I want to achieve:
[Upload a previous monthly AR narrative report from your team as markdown, or describe the structure — e.g., “Our CFO likes a 1-page executive summary first, then a table showing bucket movement, then a bullet list of top 5 risk accounts with recommended actions.”]
Here’s what makes this reference work:
The narrative always starts with the headline number (total AR balance and DSO) and compares it to prior month
It explains changes in each aging bucket with at least one concrete reason (e.g., “large seasonal order from Customer X pushed 31-60 bucket up 12%”)
It highlights accounts with repeated late payments and suggests specific actions (hold future orders, require prepayment, escalate to collections agency)
It ends with a 30-day cash collection forecast based on historical collection rates
Here’s what I need for my version / SUCCESS BRIEF:
Type of output + length: A 500-700 word executive narrative, a one-page risk heatmap table (customer name, days_past_due, amount, risk level, recommended action), and a 5-bullet action plan for the next 30 days
Recipient’s reaction: The CFO should be able to make a decision on reserve increases and collection escalation without asking follow-up questions
Does NOT sound like: Generic financial boilerplate, vague statements like “we are monitoring the situation,” or overly optimistic projections without data backing
Success means: The report is approved in one review cycle, and the CFO signs off on the proposed reserve adjustments and escalation list
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.
When you run this second prompt, expect ChatGPT to ask about seasonality, known customer disputes, and whether any major accounts have changed payment behavior in the last 30 days. It should also request your historical collection rate percentage—if you don’t have it, it will ask you to estimate. That’s a feature, not a bug: the AI is forcing you to think about the data quality before it produces a report. If your collection rate is a guess, the forecast will be a guess, and the CFO will catch it. Use this prompt to surface those gaps in your data before they become embarrassing questions in the board meeting.
Here’s the practical tip for getting the most out of both prompts: never run them cold. Always attach the actual files or paste the raw data into the chat. ChatGPT’s memory of your specific customers, terms, and payment patterns is only as good as what you feed it. If you run the collection email prompt without your CRM notes, you will get generic emails that might alienate a customer who already has a dispute open. Similarly, if you run the aging report prompt without prior month data, the trend analysis will be meaningless. The prompts are designed to ask for these files—so prepare them in advance, and you’ll get output that’s genuinely board-ready.
Finally, start small. Pick your top 10 delinquent accounts and run the first prompt on just those. Review the emails, adjust the tone, and see how your sales reps react. Once you’re comfortable, scale to the full portfolio. For the aging report, run it in parallel with your manual process for one month. Compare the narratives side by side—if ChatGPT’s version misses a nuance you caught manually, add that nuance to the reference section of the prompt. Over two or three iterations, you’ll build a personalized AI analyst that knows your AR portfolio as well as your best human does, and you’ll reclaim hours every single week.
Published on 12 August 2026 on growwithgpt.com