How Prodigal uses AI to modernise loan collection

Managing loans and collecting payments is a massive undertaking. Major lenders often employ thousands of people just to handle phone calls, emails, and messages with borrowers. As these companies take on more loans, their call centres must grow proportionally, making it increasingly challenging to deliver the same quality of service to every customer, especially when operating across different countries and communication channels.

But the real challenge goes beyond just scale. Borrowers experiencing financial distress need empathetic engagement, while lenders focus on recovery. Modern customer service expectations often clash with the industry’s aggressive tactics.

Shantanu Gangal and Sangram Raje founded Prodigal in 2018 to bridge this gap. The IIT Bombay alumni combined technical and industry expertise. Raje led the US equities desk at Tower Research, a high-frequency trading firm, while Gangal developed financial services expertise at Blackstone and Fundbox, working directly with banks and lenders. They recognised that AI could finally enable empathy-driven customer service at scale.

“This sector hadn’t seen meaningful innovation for 30-40 years, since the 1990s when the industry began implementing remote work and distributed contact centres,” Gangal explains.

The solution

California-based Prodigal is an agentic AI platform for loan servicing and debt collection that autonomously manages customer interactions across voice, SMS, and email. The platform analyses payment history, communication patterns, and behavioural data to generate customised collection strategies and predictive insights.

The Prodigal Intelligence Engine (PIE) integrates data from CRM systems, lending platforms, dialers, email systems, and external sources into a unified store. This gives AI agents complete context for every customer interaction, enabling personalised communication and automated workflows.

ProAgent autonomously handles collection conversations 24/7 while maintaining regulatory compliance. It manages routine interactions like payment reminders and follow-ups, allowing human agents to focus on complex cases.

Prodigal uses a model-agnostic architecture with Claude Sonnet 3.5, GPT-4, and Llama 3. Smaller, fine-tuned models handle repetitive tasks, such as greetings and payments, while larger models manage conversation flow and reasoning. The system uses sound cues and conversation context to know when to listen, pause, or speak.

The platform’s key differentiator is real-time negotiation. “Our agents don’t follow static scripts; they reason through each borrower’s situation and calculate personalised offers on the call itself,” Gangal explains. If a borrower proposes a different amount, the agent recalculates based on affordability models, client policies, and past behaviour to propose workable plans.

To ensure compliance, Prodigal uses engineering and rule-based coding. Client policies are automatically enforced; for example, if a policy permits only 12-month payment plans, the system rejects longer durations, combining conversational flexibility with reliable decision-making.

The three-tier architecture includes: foundation models evaluated for performance and cost; a middle layer encoding compliance and collections expertise; and a top layer with customer-specific customisation and isolated data.

An evaluation framework tests thousands of conversation scenarios for tone, accuracy, latency, and compliance before deployment. In production, agents can be fine-tuned within minutes as new data arrives.

Supporting products include ProPay (self-service payment portal), ProInsight (interaction analysis across 200+ parameters), ScoreGenie (automated performance scoring), ProCollect (digital strategy orchestration), ProScore (behavioural predictions), ProNotes (real-time documentation), and ProAssist (live agent coaching).

“Our competitive edge comes from proprietary first-party data with billions of interactions accumulated over seven years, growing by over a billion interactions annually,” Gangal says.

Prodigal claims clients have seen up to 12% payment increases, 10% labour cost savings, and 8% increases in right-party contacts. The startup serves loan servicers and collection agencies in consumer finance.

Clients and the market

The 64-member startup competes with companies like Sierra.ai in the AI-powered collections space. Its customer base includes over 100 leading financial institutions across multiple asset classes, including BNPL companies, auto finance, healthcare revenue cycle management, and debt buyers. Its clients include Prosper Marketplace, Resurgent, Halsted Financial Services, and FirstSource, an Indian BPM company.

The company operates on a usage-based pricing model, charging customers based on metered interactions on their platform. With operations in Mountain View, California, and Bengaluru and Mumbai in India, the startup is capitalising on the use of AI in the global debt settlement market.

This market was valued at $9.83 billion in 2024 and is projected to grow from $10.46 billion in 2025 to about $18.28 billion by 2034, registering a CAGR of 6.4% during the period, according to a report by Precedence Research.

The report attributes this growth to rising consumer debt levels across credit cards, personal loans, student loans, and medical loans, which are fueling demand for effective debt resolution and collection solutions.

Funding and growth

Prodigal has raised $14 million in total equity capital to date. The company secured an initial $2 million seed round from Y Combinator and Accel, followed by a $12 million Series A round led by Accel and Menlo Ventures, supported by other angel investors.

These funds will be used to expand its team across AI engineers, machine learning engineers, and forward-deployed engineers in 2026. Its current improvement areas centre on the voice agents’ natural conversation flow, contextual understanding, and replicating human pauses and instincts in dialogue. “These may sound like small nuances, but they make an outsized difference in how trust and empathy are built over a call,” Gangal says.

Prodigal’s long-term aim is to transform how lenders understand and interact with consumers through AI-driven personalisation at scale. “Being the most consumer-friendly lender will determine market winners over the next 50 years,” Gangal notes.