Reactivating At-Risk Borrowers: Awaaz AI Voice Calls Increase On-Time Repayment by 6.4% for Tala in Mexico
Tala
Key Statistics
- Pilot customers: 7,438 (3,624 called vs. 3,814 control group)
- Period: Mar 28, 2026 – Apr 15, 2026 (3 weeks)
- Engagement Metrics
- Average Call Duration: 67 seconds (reached calls)
- Language: Spanish
- On-time repayment lift, full population: +6.4% more than the control group
- On-time repayment lift, long-term customers with 10+ loans: +7.8% more than the control group
- Increase in app reinstallation rate: +2.2% capped at loan due date (more than the control group); +5.6% uncapped post loan due date (more than the control group)
- Scale: Tala approved the campaign for 100% of its ~7,500 DPD minus-7 customers/mo
About Tala
Tala is the AI-native credit infrastructure that connects global capital to the global majority. We combine proprietary risk intelligence with an expanding network of capital and distribution partners to power credit access at scale. To date, Tala has distributed more than $9 billion in capital to more than 14 million customers across Africa, Latin America, and Asia, building the definitive behavioral dataset on thin-file borrowers in emerging markets. Tala is now converting that foundation into shared infrastructure that partners, capital providers, and ecosystems can build on.
Client’s Challenge
Tala serves customers across Mexico with mobile-first, data-driven lending. In Tala’s internal convention, “DPD minus-7” refers to customers seven days before their loan’s due date, which is a pre-due reminder window. Tala’s loan terms range from 15 to 61 days, so this isn’t tied to a fixed monthly cycle. By the time a customer reaches DPD minus-7, some have already uninstalled the Tala app at some earlier point in the loan term, and Tala’s data shows that uninstalling is itself a strong risk signal: customers who have uninstalled are roughly twice as likely to default on their current loan as those who keep the app installed. Losing the app also cuts off Tala’s primary channel for reminders, payment links, and support, making it harder to bring the customer current and increasing the risk that a short delay turns into a longer-term default.
The important slice of this group was customers with 10 or more loans with Tala on record – borrowers with the longest relationship with Tala, and the most to lose from a broken channel. Tala needed a way to re-engage uninstalled, pre-due customers (DPD minus-7) directly by phone, in Spanish, at a volume and cost that field or agent-based outreach couldn’t match.
Solution
Awaaz AI deployed a Spanish-language conversational voice AI agent to call DPD minus-7 customers who had uninstalled the Tala app, reminding them of their upcoming payment and encouraging them to reinstall. To let Tala measure true causal impact rather than a seasonal or self-selection effect, the pilot was run as a randomized test:
- 50% of eligible customers were called daily, Monday-Saturday, with the remaining 50% held out as a control group
- The campaign ran for nearly three weeks from March 28, 2026 to April 15, 2026, giving a large enough sample (7,438 customers) to detect statistically significant effects even on a rare outcome like reinstallation
- Calls combined a payment reminder with an explicit reinstallation prompt, so the agent could address both the repayment risk and the channel-loss risk in a single conversation.
Methodology Note: The campaign and primary analysis were defined around Monday-Saturday calling days; Sundays were excluded because no customers were eligible for campaign exposure on those days.
Awaaz AI’s Impact
Across the full population, customers who received a call had 6.4% higher repayment rate than the control group, which is a highly significant effect on Tala’s primary metric. The effect was even stronger where it mattered most: among the long-tenured customers with 10 or more loans on record, the called group’s on-time repayment rate rose by 7.8%.
Reinstallation rate also rose by 5.6% on an uncapped basis (including the days after due date), and 2.2% when reinstalls are capped at the loan due date, a significant result given how infrequent reinstallation is within a short window.
On the strength of these results, the Tala team confirmed that the campaign was a success and Tala approved scaling it immediately, rolling it out from a 50% test to 100% of the eligible DPD minus-7 customers.
“What stood out to us with Awaaz AI wasn't just the results. It was how natural the conversations felt. Our customers were talking to a voice agent in Spanish that understood context, stayed patient with people who were behind on payments, and the conversation didn't feel scripted or robotic. That's a hard thing to get right, especially with a population that's already stressed about money. The DPD minus-7 results confirmed what we hoped voice AI could do for us: reach uninstalled, at-risk customers, and move the needle on repayment where we needed it most.
The support on the Awaaz side matched that quality. Whenever we flagged an issue, a script tweak, a voicemail detection edge case, a call we wanted reviewed - their team turned it around quickly and kept us in the loop the whole way through. That responsiveness gave us the confidence to keep scaling. It's the combination of a genuinely good customer experience and a team that's easy to work with that made this an easy call to scale to our full DPD minus-7 population.”
Bethy Hardeman
Senior Director & Global Product Marketing
