AR collections is a workflow where judgment and data work together. A credit manager needs to know which customers are most likely to pay late, which invoices are worth escalating now versus in two weeks, and which disputes are blocking payment on otherwise clean accounts.
Historically, that prioritization came from reading the aging report and applying experience. Experienced credit managers develop an intuition about which customers pay slowly, which are chronically late, and which disputes are real versus pretextual.
AI builds the prioritization model from actual payment history. It scores every open invoice, surfaces the highest risk accounts at the top of the queue, and ensures that the credit team's outreach effort goes where it has the highest return. The intuition of an experienced credit manager is still valuable. AI makes it more precise.
Why AR Collections Is Still Largely Manual
The standard collections workflow: open the AR aging report, sort by amount or days overdue, work through the list from the top, make calls or send emails, log the activity, and repeat.
The limitations of this approach:
- Sorting by amount prioritizes large balances, but a large balance from a reliable customer may need less attention than a smaller balance from a chronic late payer
- Sorting by days overdue treats a customer who is 5 days late for the first time the same as a customer who is routinely 30 days late
- Dispute tracking lives in email threads and spreadsheets, not in a structured workflow that connects to the AR aging
- At scale, hundreds or thousands of open invoices, manual prioritization consistently misses the accounts that are quietly building into collection problems
Where AI Creates Real Leverage
Payment Probability Scoring
AI builds a payment probability score for each open invoice based on actual payment history. The model considers:
- The customer's historical payment behavior: average days to pay, trend over the trailing 12 months, any pattern of deterioration
- Invoice amount relative to the customer's typical payment behavior: large invoices from this customer pay at a different rate than small ones
- Days outstanding relative to contracted payment terms: 15 days past due from a customer who typically pays at 10 days past due is not the same risk as 15 days past due from a customer who typically pays on time
- Industry and segment patterns where sufficient data is available
The output is an invoice level probability score that updates daily as payment behavior evolves. The aging report is replaced by a risk ranked queue.
Prioritized Collections Queue
Instead of working top to bottom through the aging report by amount or days overdue, the credit team works from a queue ranked by payment risk and expected cash impact. High balance, high-risk invoices rise to the top regardless of invoice date. Low balance invoices from customers with reliable payment histories fall to the bottom until they age further.
For a team managing 500 open invoices, this is the difference between reviewing 500 entries manually and focusing intensive outreach on the 30 that actually need it today.
First-Draft Collection Communications
AI drafts a starting point outreach email or call script for each high priority account, customized by:
- Outstanding balance and specific invoice details
- Number of days overdue
- Customer payment history context
- Whether prior outreach has occurred and what was communicated
The credit manager reviews the draft, adjusts the tone as needed for the customer relationship, and sends. The starting point replaces the blank page problem and ensures that every high priority account receives timely outreach rather than waiting for the credit manager to get to them in queue order.
Dispute Identification and Routing
AI identifies invoices with a history of dispute like prior email challenges, deductions applied to prior payments, PO reference mismatches and routes them to the appropriate resolution team rather than keeping them in the standard collections queue. A disputed invoice should not consume credit team outreach effort. It needs a resolution workflow.
Customer-Level DSO Tracking
AI tracks DSO at the customer level, not just at the company level. Customers with deteriorating payment trends surface before they affect aggregate DSO, giving the credit team a lead indicator rather than a lagging one.
What Good AI Collections Output Looks Like
- A ranked collections queue updated every morning: top 20 highest-priority accounts with payment probability score, outstanding balance, days overdue, and a draft communication
- Dispute flag on any invoice with a history of dispute or PO mismatch
- DSO trend by customer for the trailing 12 months, with deteriorating accounts highlighted
- Concentration risk flag: customers who represent more than a defined percentage of total AR with deteriorating payment behavior
Where Human Judgment Is Irreplaceable
- Credit limit decisions. AI flags the risk; the decision to reduce a credit limit or put an account on hold involves the customer relationship, the revenue impact, and legal considerations. That is a commercial judgment.
- Dispute resolution. Determining whether a dispute is valid and negotiating a resolution requires communication, commercial judgment, and authority to settle. AI identifies the dispute; the credit or accounts receivable manager resolves it.
- Escalation decisions. The decision to refer an account to a collection agency or to legal counsel requires an assessment of the customer relationship, the recovery probability, and the cost of escalation. That is human judgment.
- Customer relationship management. The most important collections conversations with large customers or long term relationships that are experiencing temporary difficulty require experienced professionals who understand the commercial context and can make commitments that matter.
Connecting AR Collections to the Cash Flow Model
AI scored collections data feeds directly into the 13 week cash flow model and the working capital forecast. Customer level payment probability improves the accuracy of the cash receipts forecast: instead of assuming all invoices collect at their stated due date, the model reflects each customer's actual payment behavior.
When a major customer's payment probability score deteriorates because their invoices are aging past their historical pattern, the cash receipts forecast updates automatically. The CFO sees the working capital implication of collections risk before it materializes, rather than after.
Start Here
Pull the last 12 months of customer payment data from your AR system: invoice date, due date, and actual payment date for each closed invoice. Calculate the actual days-to-pay for each customer and compare it to their contracted payment terms.
Rank your current open AR by the payment behavior of the customer, not just by the size of the balance. That ranking, your first manual approximation of AI generated payment scoring tells you whether your current collections outreach sequence is actually aligned with payment risk.
If it is not, the case for AI assisted prioritization is already visible in your own data.
