01
Proactive negotiation
It doesn't settle for the first offer: it proposes, steers toward the minimum payment and raises the ask when it can.
The product · Collections negotiation agent
Saldia. Negotiates, closes and collects, in one conversation.
An AI agent that negotiates by voice, chat and WhatsApp: it understands what the debtor says, proposes options within the portfolio's policy and closes the payment promise. 24 hours a day, weekends included.
Voice · Chat · WhatsApp
24/7
#1 in the recovery ranking
Already in production
The agent
The agent and what it does.
It's not a rules-based IVR: it understands what the debtor says, negotiates and closes the payment commitment — and knows when to hand off to a person.
02
Understands context
It grasps what the debtor says and responds accordingly, not with a fixed menu of options.
03
Closes the payment promise
It agrees on amount and date, confirms the commitment and records it.
04
Escalates to a human agent
When the case calls for it, it hands off to a human with the full context of the conversation.
05
Leaves a message with third parties
If someone else picks up, it leaves a message and asks for the account holder's phone; if given one, it calls.
06
Records reported payments
If the debtor says they already paid, it takes the payment details and negotiates the remaining balance.
The platform
All integrated into Saldia.
Operations, telephony and data in one place — from the call to the report.
360 balances and debt
Centralizes balances, products and debt for each customer in a single view.
Inbound + outbound dialer
Answers and calls from the same integrated platform.
100% recording and transcription
Every call is recorded and transcribed.
Automatic dispositions
It automatically classifies the outcome of every interaction.
Reports and dashboards
Contacts, promises, campaigns and the recovery curve.
Rules per portfolio
Plans, write-offs and policies configurable per client.
Results
One bot performs like several agents.
All of September on a consumer card portfolio (anonymized), early stages. Saldia versus the contact center's average collector.
1.7×
more recovered
$29.2M vs. $16.9M for the average collector.
#1
in recovery
In the recovery-per-intervention ranking.
520
payment promises
$99.6M committed, a third of the whole portfolio.
3.7×
more promises
Than the contact center's average collector.
189 accounts paid after negotiating with the bot. The bot worked the low balances; the human collectors, the high ones.
Frequently asked questions
What people ask us before starting.
What exactly is Saldia?
An artificial intelligence agent that handles collections by voice, chat and WhatsApp. It talks to the debtor, understands the answers, proposes payment options within the portfolio's policy and closes the promise. It works 24 hours a day, weekends included.
Does it replace human collectors?
No. Saldia takes the repetitive volume —lots of small-balance accounts, especially in early-stage delinquency— and hands over to a human the cases that need one, with the full context of the conversation. The team is freed up for where it actually pays off. We go into it in bot or human collector.
What happens if someone else answers, or voicemail?
If a third party answers, it leaves the message and asks for a phone number for the account holder; if it gets one, it calls. If it reaches voicemail, it detects it, records it as such and the contact stays available for a later attempt instead of being counted as worked.
Who decides what the bot can offer?
You do. Plans, write-offs, minimums and terms are configured per portfolio, and Saldia negotiates strictly within that policy: it cannot offer anything that is not authorised. Each portfolio can have its own rules.
Is there a record of what was said?
Yes. 100% of calls are recorded and transcribed, every interaction is automatically classified by how it ended, and it all feeds the reports on contacts, promises, campaigns and the recovery curve.
Is it running with real clients?
Yes, it is in production. In September, on a consumer card portfolio, it recovered 1.7 times what the average collector did and took 520 payment promises, $99.6M committed, a third of the whole portfolio. The bot worked the low balances; the human collectors, the high ones. The details are in the Sercom case.