The challenge: CM Financiera managed its portfolio with manual processes: agents dialing number by number, no intelligent prioritization, and scalability limited by team size. There was no post-interaction analysis or unified cross-channel traceability.
The solution: in January 2026, Hypernova Labs deployed CEL-IA at CM Financiera. The virtual agent CELIA contacts, negotiates, and records payment commitments autonomously across three simultaneous channels: voice (ElevenLabs), WhatsApp (Meta Cloud API), and email (SendGrid). Implementation took 7 months.
Intelligent segmentation: CEL-IA classifies each debtor by quadrants based on payment probability, overdue amount, and contactability, applying the right strategy to each profile.
Autonomous negotiation: CELIA negotiates within each portfolio's rules — minimum amounts, installments, terms, down payments. Anything beyond policy goes to human review automatically.
12 exit guards: before sending any message, CEL-IA runs 12 chained validations: prompt injection, holder identity, debt data, pricing, off-policy commitments, and more.
Post-call analysis: every call generates a transcript that CEL-IA analyzes automatically: agreement, amount, date, installments, tone, and payment intent. The analyzer is idempotent.
Automatic recording: promises and agreements are recorded with date validation and textual evidence. Automatic follow-ups and reminders via WhatsApp.
Infrastructure: runs on Google Cloud Platform (us-central1) on Cloud Run with auto-scaling from 0 to 10 instances. Full multi-tenant isolation. With prompt caching, cost per turn dropped from USD 0.115 to USD 0.008.
75% of customers who paid did so within the first 7 days. CEL-IA doesn't replace human judgment: it reserves it for where it matters.
