CASE FOURBANK · BLUEDOCS

Document validation that
performs like 4 analysts.

How Fourbank cut credit-onboarding analysis time by 85% and lifted validation accuracy to 99.2%.

Industry

Digital bank

Region

Brazil

Duration

12 weeks

Team

BlueDocs + squad

YEAR-ONE ROI

+346%

For every R$ 1 invested, R$ 4.46 came back, measured in lower operating cost and more approvals.

FIRST-QUARTER RESULTS

−85%
average analysis time
+34%
funnel approvals
99.2%
validation accuracy
−72%
rework per document
−R$2.8M
annual op-cost
12wk
from contract to go-live

THE CHALLENGE

180,000 documents per month.
4 days to approve a client.

In 2024, Fourbank was growing 4× a year. Credit onboarding, run by 32 analysts across three shifts, became the bottleneck: each client waited 4 days on average for a decision. Competitors decided in hours.

Leadership had two obvious options. Hire more analysts (expensive, slow to train) or loosen the validation bar (risky, and a regulatory concern). They looked for a third one.

We did not just want to automate. We wanted to turn the document into a decision, without losing the auditability the regulator demands.

Larissa Mendes · CTO, Fourbank

THE SOLUTION

BlueDocs configured in 12 weeks.
White Box from day one.

bluemetrics deployed BlueDocs on Fourbank's AWS infrastructure in 12 weeks, across three sprints: discovery, customization, and stabilization. It covered 47 document types with native integration into the bank's credit core.

01

Extraction

Multimodal OCR on PDFs, images, forms, and handwritten documents with over 99% accuracy.

02

Classification

47 document types routed automatically to the correct flow, with no human intervention.

03

Validation

Real-time cross-checks against the tax authority, credit bureaus, sanction lists, and internal databases.

04

Auditability

Each AI decision records confidence per field. Uncertain items go to a human reviewer, with a full audit trail.

THE JOURNEY

12 weeks, 5 milestones.

Wk 0

Diagnostic

Mapping the 47 document types and the friction points in the credit funnel.

Wk 1-4

Discovery + setup

BlueDocs configuration, integration with the core, and ingestion of historical documents.

Wk 5-8

Customization

Training models for the uncovered types and the bank-specific validation rules.

Wk 9-11

Pilot

Parallel run: 30% of volume on BlueDocs, 70% on the manual flow, with continuous validation.

Wk 12

Go-live

100% of volume on BlueDocs. The original team moved to critical analysis and customer service.

THE IMPACT

It was not automation.
It was a new operating standard.

None of the 32 analysts were let go. 18 moved to critical analysis of high-value operations (the cases the AI routes to a person because of low confidence); 8 joined a new model-governance team; 6 went to customer service. Rework dropped 72% and onboarding NPS rose 22 points.

NEXT CHAPTERS

BlueRisk in production, BlueDecision in PoC.

In 2026 Fourbank started a BlueDecision PoC for credit pricing and moved anti-fraud to BlueRisk, reusing the integration BlueDocs had already built.

Have a similar problem?

Tell us your story in 60 minutes. We will figure out whether a configured BlueDocs solves it, or whether a custom project makes more sense.

Start assessment →