insomnia.club back to site
[ playbook · finance and accounting ]

AR collections follow-up that sounds like your team, not a dunning robot

Collections often happens in the gaps of an AR specialist's week, using the same reminder template for a reliable customer who forgot and a struggling one who is avoiding you. An agent can read each account's context, draft the right message, and keep the follow-up cadence that people let slip.

who owns it

AR lead or controller, with account managers owning customer relationships

what starts it

Invoices pass due date, a customer replies to a reminder, or a promised payment date passes

01the problem and who owns it

Generic dunning emails go out on a schedule, customers ignore them, and the real follow-up depends on someone remembering to call. Meanwhile the reason an invoice is unpaid is often sitting in an email thread: a missing PO number, a disputed line, a new AP contact. Until someone reads that thread, the invoice just ages.

Finance owns days sales outstanding and cash, but the customer relationship belongs to sales or account management, who worry that aggressive collections will damage renewals. Both sides need to see the same picture.

02what the AI does, step by step

  1. Build the worklist dailyThe agent pulls open invoices from the accounting system and ranks them by amount, days overdue, customer payment history, and any upcoming renewal or order, so the largest and riskiest balances surface first.
  2. Read the account contextFor each customer it reviews the recent email threads, prior promises to pay, open disputes, and notes in the CRM, then summarizes why the invoice is likely unpaid, such as missing PO, wrong billing contact, or disputed line.
  3. Draft the right follow-upMessages are drafted in your team's voice, tailored to the situation: a resend with the PO number attached, a question to a new AP contact, or a firmer note for repeat late payers. Each references the specific invoices, amounts, and due dates.
  4. Capture replies as structured dataWhen a customer answers, the agent extracts promised payment dates, disputes, and requests for documents, and records them on the account so the next action is scheduled automatically.
  5. Route disputes and escalationsBilling disputes go to whoever can fix the invoice; relationship-sensitive accounts go to the account manager with a summary; balances past your escalation threshold go to the controller with a recommended next step.
  6. Report the state of receivablesA weekly summary shows aging movement, promises kept and broken, disputes by cause, and accounts needing a call from a person.

03systems it connects to

04human checkpoints

05what to measure

06risks and guardrails

07build vs buy

AR automation tools such as Chaser, Upflow, Versapay, and the collections modules in larger ERPs handle reminder cadences, portals, and aging dashboards. For straightforward B2B receivables, those cover most needs.

A custom agent is worth building when the reason invoices go unpaid lives in email threads and CRM notes those tools do not read, or when collections must be coordinated closely with account management because receivables and renewals are tied together.

Browse every finance and accounting playbook or the full library.

want this running in your business?

We can connect your billing system, AR mailbox, and CRM, and run the agent on your current aging report so you can judge the drafts and priorities before anything is sent.

See how we deliver it: ai agent development.

book a call drop your number

info@insomnia.club