1. Discovery & Requirements Document
Agentic AI Deployment for Post-Disruption Customer Recovery
| Client | SkyWave Airlines (Illustrative Enterprise) |
| Engagement Type | Agentic AI Implementation — Customer Experience Transformation |
| Analyst | Portfolio Case Study | Agentic Product Analyst Role |
| Date | Q2 2025 |
| Document Status | Final — v1.0 |
1. Executive Summary
SkyWave Airlines operates over 1,200 daily flights serving 85 destinations across North America and Europe. During Irregular Operations (IROPS) events — which include weather disruptions, mechanical delays, crew shortages, and air traffic control holds — the airline's contact center experiences a 400-600% surge in inbound customer contacts within 60-90 minutes. The current model relies almost entirely on human agents, resulting in extended hold times averaging 47 minutes, customer defection to self-help channels, and significant compensation costs driven by unmanaged expectations.
Business Objective
Deploy a Netomi Agentic AI solution that autonomously handles 70%+ of IROPS-related customer contacts — including flight status queries, rebooking requests, compensation claims, and proactive outreach — while maintaining CSAT above 4.2/5.0 and reducing contact center cost per contact by 55%.
2. Stakeholder Map & Discovery Participants
The following stakeholders participated in discovery workshops and requirement sessions conducted over a 3-week period:
| Stakeholder | Role | Department | Primary Interest |
| David Chen | VP Customer Experience | CX & Operations | CSAT improvement, brand reputation |
| Maria Santos | Director, Contact Center Ops | Operations | Containment rate, AHT reduction |
| Raj Patel | Head of Digital Transformation | Technology | Platform scalability, integration |
| Lisa Kwan | Chief Information Security Officer | IT/Security | Data privacy, PCI/PII compliance |
| Tom Adeyemi | Engineering Lead, APIs | Technology | Integration feasibility, SLAs |
| Sarah O'Brien | Senior Ops Analyst | Operations | Workflow accuracy, escalation logic |
3. Current State Process Analysis
3.1 IROPS Event Lifecycle (As-Is)
When an IROPS event is declared, the following sequence currently occurs:
- Flight Operations notifies Contact Center Ops via internal broadcast (avg. 12-minute lag)
- Contact center activates overflow queue routing — no automated customer outreach
- Inbound calls spike; average wait time escalates from 4 min to 47 min within 90 minutes
- Agents manually look up PNR in Sabre GDS, cross-reference rebooking availability, and offer options
- Compensation vouchers issued manually via legacy ticketing system (avg. 8 min per case)
- No automated follow-up — customers must call back to confirm rebooking
3.2 Identified Pain Points
| Pain Point | Impact Severity | Quantified Impact |
| No proactive outreach during IROPS | Critical | 80% of contacts are reactive; $2.1M annual cost |
| Manual PNR lookup and rebooking | High | Avg. 11 min per contact; 60% of handle time |
| No self-service compensation claim | High | 15,000 manual cases/year; $420K in labor |
| Inconsistent agent guidance | Medium | 14% error rate in compensation calculation |
| No omni-channel IROPS support | Medium | Chat/email delayed 4+ hours during IROPS |
4. Functional Requirements
4.1 Core Agent Capabilities
- Flight status inquiry and real-time delay/cancellation notification (inbound and proactive)
- Autonomous rebooking: offer alternatives, confirm selection, issue new boarding pass via email/SMS
- Compensation claim processing: eligibility check, voucher issuance, meal/hotel authorization
- Baggage status inquiry with proactive delay notification
- Escalation to human agent with full context transfer (zero repeat information from customer)
- Multi-channel coverage: voice IVR integration, webchat, mobile app chat, SMS, WhatsApp
4.2 Non-Functional Requirements
| Category | Requirement | Acceptance Threshold |
| Availability | System uptime during IROPS events | 99.95% SLA |
| Response Latency | Agent response time per turn | < 2.5 seconds P95 |
| Scalability | Concurrent session handling | 10,000+ simultaneous sessions |
| Security | PII data handling & encryption | AES-256 at rest, TLS 1.3 in transit |
| Compliance | Data residency & retention | US-only; 90-day logs, 7-year audit trail |
| Accuracy | Intent recognition accuracy | >= 92% across top 20 intents |
5. KPI Framework & Success Criteria
The following KPIs will govern the success of the Agentic AI deployment, measured at 30, 60, and 90 days post-launch:
| KPI | Baseline | Target | Measurement Method |
| AI Containment Rate | 0% | >= 70% | % contacts resolved without human transfer |
| CSAT Score | 3.8 / 5.0 | >= 4.2 / 5.0 | Post-interaction IVR/email survey |
| Average Handle Time | 11 min | < 4 min | Contact center platform reporting |
| IROPS Proactive Outreach | 0% | >= 85% of impacted pax notified | Event trigger logs vs. outreach logs |
| Cost per Contact | $14.20 | < $6.40 | Blended cost model (labor + platform) |
| Escalation Rate | 100% | <= 30% | % contacts transferred to human agent |
| First Contact Resolution | 61% | >= 88% | % contacts resolved in single interaction |
6. Constraints & Risk Register
| Risk | Likelihood | Impact | Mitigation |
| GDS API rate limits during peak IROPS | High | High | Async caching layer + circuit breaker |
| Low initial intent training data | Medium | High | Import 6 months of call transcripts |
| Customer distrust of AI during disruption | Medium | Medium | Clear AI disclosure + easy escalation |
| Regulatory changes to compensation rules | Low | High | Dynamic policy config in Netomi admin |
Document prepared as part of an Agentic AI solution design portfolio case study for enterprise customer experience transformation.