Voice AI for Outbound Debt Collection

A telecom provider needed to contact a high volume of customers with overdue balances without scaling up a human calling team, while staying compliant and capturing payment on the call itself.

1. 01
   Lead ingestion

   Overdue account lists land via FTP on a schedule; an automation reads the file, parses it, and upserts records into a Supabase table, avoiding duplicates on re runs.

2. 02
   Call scheduling engine

   A cron based scheduler checks each lead's status flags (active today, do not contact, time window) and batches outbound calls within permitted calling hours, with logic to skip weekends and public holidays.

3. 03
   Conversational agent

   An ElevenLabs voice agent places the call. It runs a staged conversation script with different stages corresponding to different points in the collections process (e.g. early reminder vs. formal notice vs. active enforcement), each with its own tone and required disclosures.

4. 04
   Voicemail detection

   The system detects whether a human answered or the call went to voicemail, and branches accordingly, leaving a compliant voicemail message versus continuing a live conversation.

5. 05
   SMS fallback

   For unanswered calls or as a supplement, an automated SMS is sent with a secure payment link.

6. 06
   In call payment collection

   For customers who want to pay during the call, the agent hands off to a DTMF (keypad) payment flow integrated with a payment processor, so a card payment can be captured without a human agent.

7. 07
   Multi brand routing

   The same underlying system serves multiple brands from one platform, each with its own script, voice, and caller ID, routed by which brand's lead the call belongs to.

8. 08
   Post call logging

   After each call, a webhook receives the call transcript and outcome, logs it against the lead record, and updates the account's status (contacted, promised to pay, disputed, escalated, etc.) for reporting and the next scheduling cycle.

Originally built for the telecom sector, the same architecture applies to any team facing a similar volume, compliance, or coordination problem. The stack (ElevenLabs, Twilio, Twilio Pay, and others) is stable, self hosted where it matters, and designed to be operated by a small team without a platform crew.

• voice AI for debt collection
• AI outbound calling agent
• automated payment collection voice AI
• ElevenLabs Twilio debt collection

• ElevenLabsSpeech synthesis for the voice agent. Chosen for latency and voice quality on production calls.
• TwilioTelephony carrier. Handles inbound and outbound PSTN, SMS, and recording.
• Twilio PayPCI compliant DTMF card capture during a live call, so payment happens without a human on the line.
• BraintreePayment processor for card capture and reconciliation.
• n8nWorkflow runtime and orchestration. Every scheduled job, webhook, and branch lives here so the logic is version controlled and observable.
• SupabasePostgres, auth, and storage. Holds the source of truth for records the workflows read and write.
• FTP ingestionReads scheduled feed drops from partners into the pipeline.
• Cron schedulingScheduling for batch jobs, reconciliation, and daily rollups.

If I rebuilt this today I would move the call scheduling logic out of cron and into a queue backed worker earlier. It scales cleaner as call volume grows and makes retries and rate limit handling a lot less painful.

How an AI voice agent handles outbound debt collection calls end to end. A deep technical walkthrough post, or a shorter explainer on why voicemail detection matters in voice AI.