Altur Already Collected $6 Million in Debt Using Nothing But AI Voices

The Macro: Debt Collection Is a $20 Billion Industry Running on Phone Calls

I find debt collection fascinating as a market because it is enormous, deeply unpopular, and technologically stuck in 2005. The US debt collection industry generates roughly $20 billion in annual revenue. Globally the number is much larger. And the core workflow has barely changed in decades: a human sits in a call center, dials a number, reads from a script, tries to negotiate a payment, and moves on to the next call. The hit rate on outbound collection calls is somewhere between 5% and 15%. Most calls go to voicemail. Most voicemails get ignored.

The incumbents in collection software are companies like FICO (which sells debt recovery optimization), Experian’s PowerCurve, and a long tail of legacy platforms like Latitude by Genesys and Totality from C&R Software. These tools handle case management and workflow routing but the actual collection still requires a human on the phone. Labor costs dominate the economics of every collection operation.

Voice AI has been coming for this market for years but the early attempts were terrible. Robocalls with rigid decision trees that could not handle anything beyond yes-or-no questions. Consumers learned to hang up immediately. Regulators responded with increasingly strict rules around automated calls. The TCPA in the US and similar regulations globally created legal landmines for anyone deploying automated voice systems in collections.

What changed is that LLM-powered voice agents can now have actual conversations. They can negotiate. They can respond to objections. They can adjust payment terms. The gap between “robocall” and “AI agent that sounds human and can reason about a conversation” is the entire opportunity.

The Micro: Mexico City, WhatsApp, and $6 Million in Six Months

Altur builds autonomous voice and text AI agents that call debtors, negotiate payment plans, and follow up via WhatsApp. The agents run on proprietary telephony infrastructure designed for low latency and predictable costs.

Luis Olave is the CEO and Pedro Fernandez is the CTO. They are based in Mexico City and went through Y Combinator’s Summer 2025 batch. The company was founded in 2023, which means they had two years of development before joining YC. That matters because voice AI for regulated industries is not something you build in a weekend hackathon.

The headline number is $6 million collected using AI agents alone in six months of operation in Mexico. That is not a projection or a simulation. That is money that moved from debtor accounts to creditor accounts through conversations conducted entirely by artificial intelligence. I do not know of another voice AI company in the collections space that has published a number like that.

The product works in three stages. First, the voice agent calls the debtor using natural speech patterns. This is not a robocall with canned responses. The agent can negotiate, handle objections, and discuss payment terms in real time. Second, after the call, the system sends automatic WhatsApp follow-ups. In Latin America, WhatsApp is the dominant communication channel and using it for follow-ups dramatically increases engagement compared to email or SMS. Third, a compliance layer monitors every interaction to ensure regulatory adherence.

Their pricing model offers two options. Per-minute usage-based pricing works like a standard SaaS model. Outcome-based contracts are more interesting because Altur essentially acts as the debt collector itself, taking a percentage of what it recovers. That second model aligns incentives perfectly and gives clients zero downside risk.

The competitive landscape in AI-powered collections includes Skit.ai, which raised $50 million and operates primarily in the US and India. TrueAccord focuses on digital-first collections using email and SMS but not voice. Prodigal AI does call analytics for collections but is not making the calls themselves. Altur is differentiated by the combination of voice agents, WhatsApp integration, and proprietary telephony in a market (Latin America) where the other players have minimal presence.

Mexico is a smart beachhead. The country has a large consumer credit market, less regulatory complexity than the US around automated calls, and WhatsApp penetration above 90%. Building a voice AI collections system that works in Mexico first and then expanding to other Latin American markets and eventually the US is a more viable path than trying to crack the US market first with all its TCPA baggage.

The Verdict

Altur has the most compelling proof point I have seen from a voice AI startup in any vertical. Six million dollars collected. Not leads generated, not calls made, not demos booked. Actual revenue recovered for clients. That cuts through every objection about whether AI agents can handle the nuance and emotional complexity of debt collection conversations.

The risk is regulatory. Debt collection is one of the most heavily regulated industries globally and the rules are getting stricter, not looser. AI-specific regulations around automated calling and disclosure requirements are still being written in most jurisdictions. One unfavorable ruling could force significant product changes.

In thirty days, I want to see the monthly collection run rate and whether it is accelerating. Sixty days, the question is geographic expansion. Are they signing banks in Brazil or Colombia? Ninety days, I want to know if they have started conversations with US clients, because that market is ten times the size of Mexico but ten times more regulated. The product works. The numbers prove it. The question is how fast they can scale before Skit.ai or a new entrant catches up.