Serasa Experian × PagVeloz · Banking & M&A · 2024

A unified checkout for securely paying multiple debts

Technical and operational integration of a payment institution that transformed a fragmented journey into a traceable, auditable, and scalable financial platform.

135 mipayments in 12 months

80% → 22%journey abandonment

≈ 1 minto pay 4–5 debts

0measured reconciliation errors

From negotiation to payment: the strategic gap

Serasa helped consumers negotiate debts, but payment happened outside its ecosystem. Customers with four or five debts had to copy each bill, switch to a banking app, and repeat the process. This created friction, limited visibility, and exposure to fake bills.

Select debts

Generate one payment

Apply antifraud checks

Orchestrate creditors

Reconcile ledger

Confirm to customer

Central decision: treat the checkout as a regulated financial operation with a ledger, idempotency, end-to-end tracking, reconciliation, controlled reversals, auditability, and human approval for sensitive adjustments.

What I personally led

Formal responsibility

SRE, DevOps, infrastructure, antifraud, and the digital wallet organization.

Acquisition integration

Technical due diligence, integration plan, target architecture, infrastructure evolution, and operational readiness.

Executive coordination

Five Project Leaders, two PagVeloz Senior Project Leaders, and alignment with Product, Finance, Legal, Security, and Compliance.

Solution architecture

Each transaction received an idempotency key, flow identifier, and tracking ID. This prevented duplicates, enabled complete journey reconstruction, supported settlement reconciliation, and controlled reversals.

CheckoutSingle Pix or boleto payment

APIs + Goprivate microservices

Kafkaasynchronous events and status

Ledger + RDSbalances, settlement, and audit

CreditorsPagVeloz and Autbank

AntifraudSerasa

Redisstate and performance

Databricksdaily reconciliation

Bedrock + RAGassisted investigation

Observabilityend-to-end tracking

Infrastructure

EKS, private VPCs, VPN, managed Kafka, Redis/ElastiCache, RDS, API Gateway, and WAF.

Security

PCI DSS, LGPD, KBA, segregation of duties, encryption, audit trails, and protection against account takeover and MITM attacks.

Settlement

Sweep account or identified corporate account at PagVeloz, depending on each creditor model.

AI-assisted reconciliation with human control

Amazon Bedrock agents with RAG correlated checkout, ledger, idempotency, settlements, refunds, and daily events. They investigated missing payments, duplicate risks, broken sequences, and settlement discrepancies. No agent was authorized to execute financial adjustments.

1 · Agent investigates

Reconstructs the journey through the tracking ID, consults validated rules, and gathers evidence.

→

2 · Recommends

Explains the probable cause, confidence level, and proposed reconciliation action.

→

3 · Human decides

A financial manager approves, modifies, or rejects the recommendation before any adjustment.

Execution and rollout

Approximately 6 months

Checkout delivery, financial integration, governance, and operational readiness.

Azure → AWS

Transparent PagVeloz migration without disrupting the existing operation.

10 days

Blue-green rollout with employees, followed by a canary based on customer groups until reaching 100% of the target audience.

Measurable impact

Abandonment80% → 22% · −58 p.p.

Fake-bill complaintsdown nearly 80%

Fraudapproximately 10% reduction

135 million

Debt-bill payments processed in the following year.

500,000 → 1.5 million

Premium Antifraude customers within five months; the checkout was one of the contributing channels.

Zero measured errors

AI-assisted reconciliation with human validation in the monitored production flow.

Lesson: the acquisition created value by combining regulatory and payment capabilities with customer reach, antifraud intelligence, data, and platform assets. AI accelerated investigations, while financial accountability remained human, traceable, and auditable.