01
Fixed cost per contact
Dialing, waiting, talking and logging costs the same regardless of the balance.
Collections operations
It's the bucket where human collectors get the least return, and where good automation shows the most. It's also where automating badly is easiest. Here's what to look at before deciding.
The problem
In early-stage delinquency the debtor doesn't yet think of themselves as delinquent. Often they missed the date, changed cards or simply forgot. The balance is small, willingness to pay is high, and the window to settle it before it escalates is short.
The problem isn't the conversation: it's the arithmetic. A collector costs the same on a low-balance account as on a high-balance one, but recovers far less. And because there are so many accounts, the team spends the day on the bucket that pays the least while the large accounts wait.
01
Dialing, waiting, talking and logging costs the same regardless of the balance.
02
Every billing cycle refills the bucket with new accounts.
03
The more time passes, the lower the odds of recovery and the more expensive it gets.
What to automate
The line isn't drawn by technology: it's drawn by how much judgment each case needs. These conversations automate well because the desired outcome is clear and the negotiating room is set in advance by the portfolio's policy.
Reminders and clean-up on low-balance accounts. Offering a plan within policy. Taking a payment promise and logging it. Resending the link or the receipt. Trying again at different hours. Leaving a message with a third party and calling the number they give you.
Complaints and disputes over the balance. Large accounts, where a badly offered write-off costs money. Situations involving vulnerability or conflict. Any case that falls outside policy and needs an approval.
The practical rule: if settling the case requires an exception to the portfolio's policy, it isn't a case to automate — it's a case to hand off. An agent that doesn't know when to pass the ball does more damage than one that handles fewer calls.
How it's done
An IVR with a menu of options is no use in this bucket: the debtor says things that aren't on the menu. What's needed is an agent that understands what it's told and answers within policy.
Propose, steer back to the minimum payment, and raise the ask when the debtor gives room.
Agree on amount and date, leave the promise on record and send the receipt.
Voice, chat and WhatsApp, at the hour the person can actually take a call.
When it escalates, the collector gets the whole conversation, not a blank record.
Plans and write-offs configurable per portfolio, with approvals by role.
Recording, transcript and disposition for every interaction, so it can be audited.
How to measure it
The most common mistake is measuring the bot against itself — "it closed 200 promises" says nothing on its own. The comparison that matters is against what your team does today, on the same bucket and over the same period.
Look at number of promises, amount promised and average ticket, all three together. A lower ticket isn't a bad sign if volume makes up for it: it means the agent is working the small accounts, which is exactly what it was put there for. And don't drop the kept-promise rate: a promise that isn't paid isn't a recovery.
In the Sercom pilot we measured exactly this way, on a retail card portfolio and against the contact center's average collector. Across both buckets of that portfolio — early and late — the agent closed 295 promises for $49.5M in two weeks: 4.6× more promises than the average collector, with a lower ticket.
Let's talk
In 30 minutes we can see whether your bucket is worth automating and what to measure.
Book a call