SteadyPay scales AI-powered outbound collections calls

More calls, more financial clarity

SteadyPay's white-label embedded finance solution supports end users borrowing small amounts to bridge temporary gaps in income. With high transaction volumes, loans varying in size, and some repayments triggered automatically, SteadyPay needed a scalable way to ensure every customer received timely, relevant communications about their account. The goal was to reach customers with outstanding balances proactively and support them in committing to a clear repayment plan.

The team hypothesised that a phone call would have a higher rate of success in communicating deadlines and securing commitments, and they wanted a solution that would let them operate this initiative at scale. So, SteadyPay looked for an AI agent to operationalise outbound Voice calls for the growing group of customers that met the requirements. 

Gradient Labs’ AI Voice agent was chosen for its proven ability to handle complex calls without latency, its natural tone, and its unfailing adherence to the strict compliance standards required of financial institutions. 

SteadyPay began with a short internal-only testing period to finesse the identity verification process and the AI agent’s tone, then deployed live to customers in controlled batches, scaling over four weeks to 33,000 calls.

Precise enough for finance, without sacrificing human empathy

Loan repayment calls are nuanced and critical to get right the first time. From the beginning, SteadyPay had a clear vision of what success must look like for the AI Voice agent. First, the agent had to successfully request and secure ID verification from the customer on the phone; then, it had to navigate the conversation with empathy and minimal latency. 

Key points to make in the call included clarification, such as asking the customer if they were aware of an outstanding balance. The agent also needed to communicate payment expectations, consequences, and the individual’s options for proceeding with a payment plan. Ideally, SteadyPay also wanted the agent to secure a payment commitment date from the customer, then send that information to SteadyPay’s back office systems and update the customer’s records. 

Gradient Labs’ AI voice agent was well-equipped to handle these demands. It had already ingested examples of past calls with human agents, allowing it to quickly grasp SteadyPay’s tone of voice and policy nuances. As an early adopter of AI voice technology in regulated financial services, SteadyPay worked closely with Gradient Labs to refine the agent’s guardrails for the specific requirements of outbound collections in a compliance-sensitive environment. 

How guardrails ensure safety

Gradient Labs has a suite of financial services-specific guardrails which define the boundaries of what each AI agent can discuss, ensuring every call stays compliant and prioritises customer safety. In typical inbound support chats for a financial institution, the AI agent wouldn’t discuss financial difficulties with a customer, as legally only licensed professionals can give out financial advice. 

However, in the case of loan collection, this topic is safe, as it focuses on existing balances, payment options, expectations, and consequences. The Gradient Labs team and SteadyPay worked together to quickly teach the AI voice agent this exception and the nuances around it, achieved through a rapid period of internal guidance and testing. 

These guardrails were not only critical to get right from a compliance perspective; they were essential for delivering on empathy. Customers with outstanding loan balances can be navigating external pressures, which can factor into repayment conversations. Only with extremely precise guardrails can an AI agent handle these conversations meaningfully while staying compliant.

After a few weeks of internal testing, Gradient Labs’ AI voice agent was ready for live deployment with real customers.

For some, a phone call is all it takes

Over the course of four weeks, the Gradient Labs AI voice agent made over 33,000 outbound calls to SteadyPay customers who had a loan and had not yet responded to written communication.

SteadyPay found that if a customer picked up and passed identity verification, there was a 60% chance of meeting the AI agent’s goal: securing a payment date, or a promise from the customer that they understood their circumstance and would select a payment date. 

Some customers picked up the phone but didn’t have time to speak; in these cases, the AI agent was able to ask for a better time, schedule a callback, and initiate that callback automatically to the customer without any human oversight. This empowered SteadyPay to reach customers at their ideal times, even outside of regular working hours for their support team. 

Additionally, after a few weeks of outbound calls from Gradient Labs, SteadyPay began to see a rapid uptick in inbound calls from customers who wanted to discuss their payment options.

These customers, who were considered lapsed or “cold,” were recategorised as engaged, as they had reached out proactively after seeing the voicemail or missed call and could confirm awareness of their loan balance and payment expectations. In one month, SteadyPay saw a 20% jump in customers who moved from “cold” to engaged, simply from inbound calls. 

While the outbound call programme is still ramping, it’s clear that SteadyPay has found a way to scale one-on-one communication with customers.

This communication is critical to ensure each person has financial clarity around their loan status, can commit to a repayment plan that works for them, and can minimise missed payments, which is better for overall financial health and credit scores. It’s a win for both SteadyPay and the customers who rely on loans for financial control.