The answer: build one collections workflow, not two channels
WhatsApp and voice agents solve different parts of the same receivables workflow. WhatsApp is strong for invoice resend, payment links, short reminders, receipt collection, and quick promise-to-pay capture. Voice is stronger when urgency, confusion, or a missed promise requires a live conversation.
The operating mistake is to run them as separate campaigns. A customer should not receive a WhatsApp reminder, a voice call, and an email that disagree about the balance, payment plan, or next step. The AI layer should coordinate the case, not multiply touches.
Best-practice workflow
- Segment accounts by aging bucket, amount, customer value, dispute state, previous promise behavior, legal status, and preferred channel.
- Use WhatsApp first when the job is convenience: resend invoice, confirm balance, send link, receive proof, or capture a simple payment commitment.
- Use voice agents when the account needs explanation, identity verification, callback handling, missed-promise recovery, or first-pass dispute intake.
- Pause automation on hardship language, legal threats, sensitive complaints, identity uncertainty, unusual payment terms, and complex disputes.
- Write every outcome back to AR or CRM: promise date, promised amount, dispute reason, payment link sent, receipt received, callback requested, or escalation reason.
Governance before volume
Collections is a bad place to learn governance after launch. Before increasing volume, define approved language, quiet-hour rules, consent checks, identity checks, recording policy, supervisor review, payment-plan limits, and exception routing.
The agent should expose why it acted: account data used, policy applied, template selected, channel chosen, customer response, and resulting system update. That audit trail is what separates an operating model from an uncontrolled automation project.
Metrics that matter
- Promise-to-pay capture rate by channel and aging bucket.
- Promise kept rate, not only promise volume.
- Dispute classification rate and time to route the dispute.
- Human escalation rate by reason, agent, policy, and customer segment.
- Cash collected per compliant contact, not raw messages or call attempts.
- Writeback completeness: how often the account record matches the conversation outcome.
How SOBERAN fits
SOBERAN connects WhatsApp, voice, email, AR context, CRM records, approvals, and human review into one collections loop. The agent can handle routine reminders, invoice questions, payment links, promise capture, callbacks, and dispute intake while keeping sensitive decisions under human control.
For buyers comparing AI collections platforms, ask to see the full case history: account selection, channel decision, approved message or script, customer response, policy result, writeback field, and escalation packet.