AI Debt Collection

AI debt collection is the use of conversational artificial intelligence — typically deployed on WhatsApp or other messaging channels — to handle outbound contact, negotiation and payment routing across the collections lifecycle. It replaces the traditional cost centre of dial-and-pray call operations with structured, compliant automation that scales without proportional headcount growth.

In a modern conversational AI debt collection deployment, the AI plays three roles:

• Reach — agents message the debtor on a channel they actually answer. WhatsApp open rates exceed 90% per Meta's published business platform data, versus 10–20% for SMS and roughly 30% for outbound calls.
• Negotiate — AI handles the first three to five turns: verifies identity, presents the balance, offers pre-approved payment plans within guardrails your credit team has signed off.
• Hand-off — AI escalates to a human agent only when complexity, dispute, hardship or sentiment requires it. McKinsey's 2022 collections research finds conversational AI handles 60–80% of routine collections interactions without escalation.

The net effect is an order-of-magnitude shift in cost-to-collect economics: a single AI agent runs thousands of parallel conversations at marginal cost, while human collectors are reserved for the cases that genuinely need them.

Pre-due reminders

Three to five days before a payment is due, the AI agent sends a contextual reminder over WhatsApp — balance, due date, one-tap payment link. Customers who would have forgotten get a frictionless nudge before they ever enter the collections funnel, lowering bucket-1 inflow.

Bucket 1 (1–30 days overdue)

Soft conversational outreach. The AI verifies identity, presents the missed amount, offers an immediate payment link or a pre-approved short-term plan. Most self-cure cases close here without a single human touch — freeing collections agents for higher-value work.

Bucket 2 (30–60 days)

Structured negotiation. The AI presents tiered payment-plan options within guardrails your credit team pre-approved — deposit + instalments, settlement discount, payment holiday. Promises-to-pay are captured, scheduled and tracked to kept-promise outcomes.

Late-stage (60+ days)

AI assists the human collector rather than replacing them. It gathers documentation, schedules callbacks at debtor-preferred times, summarises prior conversations and pre-fills the case file — so the agent opens a complete record instead of starting cold.

Skip tracing assistance

When a primary contact number goes silent, AI cross-references CRM and adjacent records to surface alternative WhatsApp numbers, family contacts or workplace details — turning weeks of manual skip work into a same-day automated check.

Promise-to-pay tracking

Every promise-to-pay made over WhatsApp is captured as structured data, scheduled in the system of record and followed up automatically on the agreed date. Kept-promise rate becomes a tracked KPI rather than an estimate.

Meta's published WhatsApp Business data reports 90%+ open rates on conversational messages, against 10–20% for SMS and ~30% for outbound calls. The volume of debtors you can actually reach changes the economics before any negotiation logic runs.

Every WhatsApp message is a timestamped, persistent record. Disputes are resolved from a complete conversation log, not from a collector's recollection of a call. Regulators get a clean audit; customers get a clear receipt of every interaction.

English, French, Afrikaans and Arabic out of the box — relevant for the South African, francophone-African and Maghreb portfolios FCB.ai's clients operate.

Workflows are configurable to meet FAIS and POPIA (South Africa), the CIMA insurance code (francophone Africa) and GDPR (EU debtors) — including consent capture, contact-hour enforcement, opt-out handling and PII redaction.

The conversation, the document capture, the payment hand-off, the audit log — all live on a channel the debtor already uses every day. No app install, no portal login, no email that gets ignored.

Your legal and compliance teams approve the conversation tree — tone, payment plan ranges, escalation triggers, contact hours, language for vulnerable customers. The AI cannot say anything outside what you signed off.

Structured promise-to-pay records, payment events, conversation transcripts and outcome data flow into your existing collections system via API. No parallel database. No reconciliation overhead.

Contact rate, promise-to-pay rate, kept-promise rate, recovered amount per outreach, agent escalation rate — reported by bucket, by portfolio segment, by campaign. Optimisation is data-driven, not anecdotal.

English, French, Afrikaans and Arabic out of the box — relevant for FCB's South African, francophone-African and Maghreb client base. Switching language is a debtor preference, not an engineering project.

Every message, consent event, payment promise, escalation and opt-out is timestamped and stored. Disputes are resolved from a complete record, not from a collector's memory.

Protected debtor information, opt-out handling, contact-hour rules and data-residency controls — all enforceable at the workflow layer.

Equivalent regulations covering credit and insurance debtor interactions across the CIMA zone, supported through configurable French-language workflows.

Consent capture, right-to-erasure handling and lawful-basis tracking for any EU-resident debtor, with PII redaction in stored transcripts.

Message-level audit logs for every interaction, automatic PII redaction in transcripts shared with downstream systems, configurable data-retention policies.