An AI agent already negotiates and closes binding payment plans with debtors on live calls, in production today. A separate engine sets every offer and disclosure; the AI only voices them.
An AI collections agent runs the full arc of a collections conversation. It reaches the debtor, verifies identity, negotiates terms, closes a binding payment plan on the call, and then follows each due payment afterward. Across all of it, a separate engine sets every offer, and the AI agent voices it.
The end-to-end process
Start with contact. The AI agent reaches the debtor by phone or in text, verifies identity, and delivers the required disclosure. It works out why the debtor is reachable now and opens the conversation about repayment.
Then the negotiation. The AI agent asks what is feasible and listens to the answer. The Decision Engine returns the installment plans the account qualifies for, with minimums and conditions attached. The AI agent presents those options in plain language, answers questions, and collects a choice.
Then the close. The Decision Engine checks the choice, forms the binding plan, and issues the mandate to charge. The plan exists by the end of the same call, without a callback or a handoff between departments.
Then the follow-through. The AI agent tracks the schedule that was agreed, reminds the debtor before a due date, and reopens the conversation if a payment does not arrive. The full chain, from first contact to collected installment, runs without a person on the creditor's side for the standard cases.
One call, start to finish
Take a routine call. The debtor picks up and says they know about the balance but cannot clear it in full today.
The AI agent verifies identity and gives the required disclosure. It asks what is feasible this month. The Decision Engine returns the installment options the account qualifies for, each with its minimum and terms. The AI agent lays them out, answers a question about the first due date, and the debtor picks a plan. The Decision Engine checks the choice, forms the binding plan, and issues the mandate to charge. Before the call ends, the debtor has a schedule and a clear next step, and every decision along the way is recorded with the conditions that were checked.
What the engine decides, and what the AI agent says
The division is the whole point. The AI agent owns the conversation: understanding, tone, objection handling, and language. The Decision Engine owns the decisions: which plans apply, what the minimums are, which disclosure is required, and when to escalate to a human agent.
The same account state always returns the same set of allowable terms. The AI agent forms the sentences, and the engine sets the amounts and conditions. Because the offers come from the engine, the AI agent cannot propose a term the policy does not allow, however accommodating that term might sound. The amounts on a live call work exactly as they do in voice AI on a live call.
What it does not do on its own
An AI collections agent has a clear boundary, and that boundary is by design. It does not invent concessions, because concessions are not in the set the engine returns. It does not push past a request to stop contact, because conduct rules are part of what the engine enforces. When a debtor signals hardship, raises a dispute, or presents a case outside standard policy, the AI agent hands the conversation to a human agent with the full context: who called, what was agreed, and where the conversation paused.
That handoff keeps judgment with people. Standard plans close automatically, and the decisions that need a human stay with one.
What it connects to
An AI collections agent is only useful when it acts on real account data. It reads the account state through an integration, passes that state to the Decision Engine, and applies the offers the engine returns. The conversation is grounded in the creditor's own policy and records, so the plan it closes reflects the account in front of it.
On the phone and in text
The same AI agent works across channels. In text it holds a threaded conversation and closes a plan; on a call it recognizes speech and replies in natural tempo. The logic behind the offers is identical in both, because it lives in the engine rather than the channel. Every decision is logged, call by call, which is what makes it reviewable for compliance.
An AI collections agent built this way runs in production today, in voice and text.
Hear a call
The clearest test is your own ear. Hear a sample collections call handled by an AI agent, as a demonstration: Zowie Voice.
