Tala is AI-native credit infrastructure for the global majority. They provides short-term loans to users across Mexico. Tala was built for thin-file borrowers in informal economies, not adapted from legacy systems. Trained on more than $8B in lending performance data across 14 million customers.
Tala App customers receive an automated reminder call powered by Awaaz.ai 1 to 5 days before their loan due date. The call is sent to encourage timely repayment and reduce missed payments.
Region: Mexico Language: Spanish Agent Run time: 6 months
🇲🇽 212K+ Confirmed Identity 💵 71% gave disposition
_breakdown-8_13_2026_4_46_24_PM.png)
Assessing which user
Why manual review? Because insights like the one below can be easy for an LLM to miss and are often best identified by a human looking at the tone and context, STT Functionalities, real like BG sounds, and colloqialisms
| Deep-dive into user responses 🗣️ | Qualitative Insights 👀 | Tech Fixes / LLM Updates 👩🏻💻 |
|---|---|---|
| 1. Users took longer to respond when asked for a payment date | ||
| When the agent asked, “Can you pay by 23rd September?”, users often paused before answering. | Users took longer to respond when asked for a payment date | |
| When the agent asked, “Can you pay by 23rd September?”, users often paused before answering. | Made payment dates easier to understand. |
We updated the prompt to: “Can you pay by Wednesday, 23rd September?” | | 2. Users who couldn’t pay on time responded better to empathy before an extension was offered The original flow asked why they couldn’t pay, then offered a 3-day extension. | The order of the conversation Users were more receptive when the agent first acknowledged their situation. This made the 3-day extension feel like a response to their circumstances, rather than a generic option being offered by the system. | Added an empathy step before offering the extension. The agent now acknowledges and validates the user's situation before triggering the 3-day grace-period logic. | | 3. A simple “Yes” wasn’t always a strong payment commitment When asked to confirm whether they would pay on time, some users responded with only a one-word “Yes.” | A short confirmation doesn’t necessarily mean strong intent. A one-word response can indicate passive agreement or that the user is simply trying to move through the conversation. We needed a stronger confirmation of intent. | Added a final payment reminder to reinforce commitment.The closing now reminds users of the 10% late-payment penalty, reinforcing the importance of paying on time and prompting a more deliberate confirmation. |