Promise to pay capture with AI for lending and NBFC collection calls
Agree an exact amount, date and payment method on each collections call with an AI voice agent in natural text to speech, then remind the borrower before the promise falls due.
Build a promise to pay script
A promise to pay is a borrower's commitment to pay a set amount by a set date, and it is the main outcome a collections call aims for. PyAI captures it in a structured way: the voice agent agrees the amount, date and method, saves them through a signed webhook, and calls again the day before. Trace scores every call, and agents cost $0.08/min. A promise is logged only with a date, an amount and a method, and PyAI supplies the voice infrastructure, so the lender keeps its own loan system and rules. Credit unions, auto lenders and BNPL teams run the same flow on early stage arrears.
How promise to pay capture works
Overdue accounts come in, promises go out to the loan system, and reminders run on their own until the money lands. Each step writes a status back, so collectors see where every account stands.
Pulled from the loan system each morning.
Identity checked before any detail.
Amount, date and method.
Promise made
SavedStructured promise sent to the loan system
ReminderScheduled for the day before
No promise
Reason loggedJob loss, dispute or other
CollectorLive transfer with a whisper
Payment checked on the date and the next call set.
Promise to pay use cases for lenders
Lenders and NBFCs use the same agent at each point of the promise cycle.
Exact promise capture
Amount, date and method, nothing vague.
The agent will not accept soon or next week. It asks for a date and reads the full promise back for confirmation.
Recurring promises
Split arrears into a set of dated payments.
When the full amount is too much, the agent offers the instalment rules the lender allows and saves each date.
Day before reminder
A short call or text before each promise is due.
Reminders lift the kept rate. The agent repeats the amount and sends a payment link. Borrowers who need to move the date can do so on the same call, and the new date replaces the old one.
Broken promise follow up
A next day call when a promise is missed.
The agent asks what changed, records the reason and sets a new date or hands over to a collector.
PTP rate and kept rate
Clean data for both numbers.
Because every promise is structured, ptp collections reporting shows the right party contact rate, promises per contact and the kept rate without manual tagging.
Dispute and hardship routing
Calls that need a person go to one.
A borrower who disputes the debt or reports a hardship such as job loss is not pushed for a promise. The agent logs the reason, stops the ask and transfers to a collector with a whisper, or books a callback inside allowed hours.
Promise to pay call cost calculator
Set the call volume, call length and how long a call takes your staff to dial by hand. The total includes the agent, phone minutes and answering machine detection.
AI voice agent (Agents)$1920.00
Phone minutes$240.00
Answering machine detection$28.00
Exact promises, captured cleanly
A promise is only useful if the details are exact. Hear transcribes every call at $0.001/min, so the amount and date in the record match what the borrower said. Omni replies in about 390 ms voice to voice, which keeps borrowers talking long enough to reach a firm commitment, and Recap writes a short note for the collector.
Debt collection rules for PTP calls
Collection calls are tightly regulated. What PyAI supplies and what the lender keeps are split out below.
Build a promise to pay from your own code
Wire promise to pay into your product through Agents in one of three ways. Full docs: PyAI SDKs.
Call Agents as a plain SDK function, pass the caller's words and read back the agent's reply.
Audio and data come back ready to drop into the step that follows.
pip install pyai
Over MCP, an assistant gains Agents as a tool and runs your promise to pay inside the conversation.
You grant access once and can revoke it any time, and every call is logged.
pyai mcp add
Prototype the prompts for your promise to pay with the CLI: log in once, then voice one line per command.
Each run saves a file you can play or pipe straight into the next step.
pyai speak "..."
Promise to pay FAQ
What is a promise to pay in collections?
A promise to pay is a borrower's commitment, made on a call or in writing, to pay a specific amount by a specific date. It does not guarantee payment, but it gives the collections team a dated commitment to track, remind and forecast against, and a clear trigger for follow up if it is missed.
What is the difference between PTP rate and kept rate?
PTP rate is the share of right party contacts that end in a promise. Kept rate is the share of those promises that are actually paid. A team can have many promises and still collect little, so both numbers matter. Structured capture makes both easy to measure.
How does an AI agent capture a promise to pay?
The agent confirms identity, states the overdue amount, then asks for an exact amount, date and payment method. It reads the promise back, saves it through a signed HTTPS webhook to the loan system, and schedules a reminder the day before it falls due.
Are AI ptp collections calls compliant?
They can be when they follow the debt collection and calling rules that apply to the lender. In the US, Regulation F limits call frequency and counts voicemails as calls. Have counsel review scripts. Trace scores each finished collection call for TCPA and PII rules on any plan. Recordings and transcripts sit in encrypted storage scoped to your organization. This is not legal advice.
What does promise to pay calling cost?
PyAI Agents cost $0.08/min with no monthly fee, plus $0.01/min for a PyAI number or your own carrier. A 2 minute call is about 18 cents. Answering machine detection is free for 5,000 calls, then $0.004 per call. New accounts get $50 in credit.
Can promise to pay calls run in Spanish?
Yes. PyAI Agents speak Spanish, English, French, Hindi, German, Italian, Portuguese, Japanese, Korean, Haitian Creole, Mandarin and Arabic. A lender can set the language per borrower record, and the promise is saved to the loan system in the same structured format whichever language the borrower used on the call.