AI collections

What is AI collections?

AI collections is the use of technology that holds the conversation with a customer in arrears, rather than technology that decides who a person should call next. The distinction is what the system does when the customer answers: a dialer hands the call to an agent, an AI agent works the call itself, within limits the lender sets.

How is it different from collections automation?

Collections automation moves work towards a person: it dials, queues, sorts and sends templated messages, and a human handles the conversation. AI collections handles the conversation. They divide on one thing: whether the system can answer something it was not scripted for, which is where a real conversation starts.

What can an AI agent do on a collections call?

Confirm who it is speaking to, give whatever disclosures the market and the lender require, establish what the customer can pay, and agree terms the lender has already approved. What it may offer, when it may call and when it must stop are set by the lender before any call is placed.

Where does a collections platform fit?

A collections platform is the system of record around the conversation: the accounts, the strategy, the rules, the reporting and the audit trail. AI agents are one way conversations get held inside it. The two are complementary, and a vendor that sells one usually integrates with the other.

Is this the same as conversational AI?

Conversational AI describes the part the customer notices, which is a system that understands speech and replies in it. Collections adds three requirements on top: the agent has to keep to limits somebody set, land on an amount the customer can genuinely afford, and leave proof afterwards of what was said.

What does it change operationally?

Capacity stops being a headcount question. A team that could work the first thirty days of arrears and pass the rest to a debt collection agency can work further down the portfolio, because the cost of a conversation no longer rises with the number of them. Which accounts get worked becomes a strategy decision rather than a staffing one.

What does it not do?

It does not decide policy. The lender keeps what may be offered, to whom, and who approves it. An agent that invents an arrangement outside those limits has malfunctioned. It also does not replace judgement on the accounts that need it, which is why every serious deployment has a handover.

Who builds AI collections technology?

Three groups, and they are not interchangeable. Collections platforms adding conversational features to a system of record. General-purpose voice AI vendors pointed at collections. And products built for arrears from the start. Murphy is in the third group, holding conversations across phone, WhatsApp, SMS, RCS and email within the limits the lender sets.