01
No shifts
It answers and calls 24 hours a day, weekends included. The debtor picks up when they can, not when your operation is open.
Comparison
Bot or human collector: when each one makes sense.
We're the ones selling the bot, so let's say it up front: it doesn't replace your team. There are buckets where it wins hands down and buckets where deploying it is throwing money away. Here's the line, based on what we've seen in production.
Where the bot wins
Volume, hours and consistency.
The automated agent's advantages aren't about intelligence: they're about scale and stamina. It does the same on account number ten as on number five hundred, on a Sunday at eleven at night.
02
It doesn't tire
Conversation two hundred has the same energy as the first. In a high-volume bucket, that's half the result.
03
It stays within policy
It won't offer a write-off that doesn't apply just to close faster. What's outside policy, it doesn't offer.
04
It scales without hiring
Raising the volume of work doesn't mean adding seats, training, or waiting out a learning curve.
05
It logs everything
Recording, transcript and disposition for 100% of interactions. Auditable without sampling.
06
It pays off on small balances
It takes the bucket where cost per contact eats up whatever a human collector recovers.
Where the person wins
Judgment, exceptions and relationships.
Anything that means stepping off the script is still human territory. And it isn't a temporary limitation that more technology fixes: in many of these cases you don't want a machine deciding.
01
Large accounts
Where a badly calibrated write-off costs more than the collector's salary. That's where you want your best person.
02
Complaints and disputes
If the debtor questions the balance or says they already paid something that isn't on record, it needs investigating, not negotiating.
03
Sensitive situations
Illness, job loss, vulnerability. They call for judgment and tact, and sometimes a decision that isn't in any policy.
04
Exceptions
Any agreement that needs a supervisor's approval. The bot can spot it and hand it off, not settle it.
05
Relationship customers
When the debtor is also an active customer you don't want to lose, it's a different conversation.
06
Rare cases
Deaths, estates, fraud, identity theft. Low volume and a high risk of getting it wrong.
The numbers we have
What happened when we put them head to head.
On a retail card portfolio (anonymized), from September 1st to 14th, we compared Saldia against the contact center's average collector on the same bucket.
4.6×
more promises
295 for the bot vs. 64 for the average collector.
3.2×
more value
$49.5M vs. $15.7M for the average collector.
$167,734
bot's ticket
Lower: it works the smaller balances.
#1
in the portfolio
First by count and by amount.
Read this carefully, because it's where people oversell. The bot didn't negotiate better than a good collector: it worked more hours on the bucket that pays a person the least. The lower average ticket confirms it. Had we put the bot on the large accounts, the result would have been different — and worse. The full detail is in the Sercom case.
How to split it
It isn't about choosing: it's about assigning each bucket well.
The useful question isn't "bot or people", it's which part of the portfolio you give to each. A split that works in practice:
To the automated agent
Early-stage delinquency and low balances, mass first contact, reminders and resending links, retries at different hours, follow-up on promises already taken, and all the work outside office hours.
To the human team
High-balance accounts, anything the bot hands off, complaints, agreements that need approval, customers with an active relationship, and the cases that call for judgment.
The condition for this to work is a clean hand-off: when the bot escalates, the collector has to receive the whole conversation. If the customer has to repeat everything from scratch, you've lost the advantage and annoyed them on top of it.
Let's talk
Want to see how the split would look on your portfolio?
We'll look at your buckets and tell you honestly what we'd automate and what we wouldn't.
Book a call