How CYF's AI Analysis Revealed 42% of Contacts Concentrated in Financial Issues and Created a Roadmap to Reduce Costs by 30% | Real Case Study
In-depth analysis of customer interactions from a digital payments company identified critical patterns of frustration, operational bottlenecks, and automation opportunities that can reduce recurrence by up to 40%
High Contact Volume and Recurring Frustration
A growing digital payments company faced high service volume, call recurrence, and frustrated customers. Understanding root causes was necessary to optimize operations.
Distribution of Contact Reasons
We analyzed thousands of interactions to identify the main reasons why customers contacted support:
| Contact Reason | Percentage |
|---|---|
| Payment transfer / receipt issues | 42% |
| Bank account / banking address changes | 18% |
| Improper charges (rental, plans, devices) | 14% |
| Pix issues | 9% |
| App, biometric, and access difficulties | 8% |
| Installation, exchange, or return of devices | 6% |
| Other (cancellations, chargeback, vouchers) | 3% |
The Real Problems Behind the Numbers
AI analysis revealed three fundamental patterns that explain most customer frustration
Financial + Communication
The biggest frustration isn't the error itself, but lack of predictability. Customers receive divergent information about credit timelines, bank changes, and responsibilities.
Main cause of recurrenceRegistration + Technology
Tax ID changes, death, ownership changes generate long, non-transparent blocks. Facial biometrics is a critical abandonment point.
Accounts stopped without billingProduct + Commercial Expectation
Misalignment between what was sold and what's delivered in fees, vouchers, rental exemption, and timelines. Trust breakdown leads to early churn.
Direct impact on retentionCritical Insight
Impact on Business Indicators
CriticalTimeline of Recommended Actions
Strategy structured in three phases with quantified gains for each stage
Short Term
0-30 days
Quick Implementation Actions
- Standardize timeline messages: Create single script for bank credit, Pix, account changes
- Single script for value transfers: Unify communication about gross vs. net, fees, and loans
- Proactive app notification: "Amount sent to bank" / "Dependent on bank timeline"
- Improper charge macro: Standardized response for returns, exchanges, and cancellations
Estimated Gains:
- Recurrence reduction: 20-30%
- AHT reduction: 10-15%
- CSAT increase: +8 to +12 points
Medium Term
1-3 months
Automation and Intelligent Flows
- Automated app flows: Account change with clear checklist and visual process status
- Intelligent call triage: Automatically separate "bank problem" vs. "customer problem"
- Charges and devices center: Specialized line or queue for these topics
- Biometrics with fallback: Alternative validation after X failed attempts
Estimated Gains:
- Total financial calls reduction: 25-35%
- Involuntary churn drop: 15-20%
- Critical escalations reduction: 30-40%
Long Term
3-6 months
Structural Transformation with AI
- Predictive financial dashboard: Customer sees exactly when and why they'll receive payment
- Automatic billing audit: Detection of improper rental and device duplication
- Risk-oriented onboarding: Special attention for customers with recent tax ID changes
- Frustration prevention AI: Detects calls with high emotional risk and prioritizes resolution
Estimated Gains:
- Structural contact reduction: 35-50%
- Retention increase: +5 to +8 percentage points
- NPS improvement: +10 to +18 points
- Operational cost reduction: 20-30%
Consolidated Operational Gain
Expected impact after full roadmap implementation in 6 months
Before vs. After Implementation
| Indicator | Current Situation | After 6 Months |
|---|---|---|
| Recurrence | High | -40% |
| Average AHT | Elevated | -25% |
| Churn by frustration | Relevant | -20% |
| NPS | Unstable | +12 to +18 |
| Cost per contact | High | -30% |
Impact on Bottom Line
Main Problems and Specific Solutions
Problem: Credit Delays and Value Discrepancies
42% of contactsCustomers don't understand why the received amount differs from expected and when exactly they'll receive it. Confusion between gross amount, net, fees, advances, and loans.
Recommended Solutions:
- Create transparent calculator in app showing: Gross Amount â Deductions (fees, loans, advances) â Net Amount
- Automatic notification: "Your payment of $X was sent to bank [Bank Name] at [time]. Expected arrival: [standard bank timeline]"
- Standardized script for agents with consistent information about bank timelines
- Predictive dashboard showing when each amount will be credited
Problem: Complex Bank Account Change
18% of contactsManual, slow process without visibility. Customers with closed or changed tax IDs are blocked without quick resolution.
Recommended Solutions:
- Self-service flow in app with automatic bank data validation
- Visual checklist showing each step: Request â Validation â Approval â Activation
- For complex cases (closed tax ID, death): specialized queue with reduced SLA
- Proactive notifications at each status change
Problem: Recurring Improper Charges
14% of contactsReturned devices still charged, poorly processed cancellations, device exchange generating double charge.
Recommended Solutions:
- Automatic audit system detecting inconsistent charges (e.g., rental for device returned over 30 days ago)
- Automatic refund for clear cases identified by AI
- Charges center with dedicated queue and immediate resolution power
- Visual confirmation in app: "Your cancellation was processed. Last charge: [date]"
Problem: Blocked or Not Enabled Pix
9% of contactsPix doesn't appear on device, is blocked on multiple terminals, elevated resolution timeline (4-7 days).
Recommended Solutions:
- Automatic diagnosis in app: customer clicks "My Pix doesn't work" and system tests connectivity and configuration
- Real-time Pix enablement via app (no contact needed)
- For complex blocks: specialized technical queue with 24h SLA
- Step-by-step visual tutorial for initial setup
Problem: Facial Biometrics and App Access
8% of contactsFacial recognition failure, blocking by attempts, impossibility to change email or phone, dependence on third parties in cases of death or partner.
Recommended Solutions:
- Alternative validation method after 3 failed biometric attempts (e.g., document + selfie)
- Email/phone change via self-service with SMS/old email validation
- Expedited process for death or partner change cases (digital documentation)
- Improve facial recognition algorithm or add permanent alternative options
Powered by CYF's AI-Based Customer Interaction Analysis
Behind every insight in this case study is CYF's advanced analysis engine â trained to process audio and text interactions from real customer service conversations. We transform raw call recordings, chat transcripts, and digital touchpoints into structured data that reveals trends, customer needs, and operational bottlenecks.
Our Solution
Our solution uses state-of-the-art speech-to-text transcription plus large language models and hybrid AI pipelines to understand:
- What customers say and how they feel
- How agents respond and where gaps occur
- Patterns visible only through data-driven intelligence
This powerful combination of audio processing and natural language AI makes possible insights like 30% reduction in operational costs and hidden friction patterns, transforming everyday service interactions into strategic business results.
