What challenges did A major U.S. airline face before AI implementation?

Airlines faced rising and complex customer-service demand across voice and digital channels. Legacy IVR and rule-based chat could not handle complex workflows such as flight changes and refunds. Customer service teams had to navigate fragmented airline systems manually, slowing resolution and increasing costs.

What was the impact of the AI solution?

The implementation was completed in 39 days. The airline reached 91% first-contact resolution, a 21% improvement over baseline. It also increased customer self-service adoption and achieved 92% customer satisfaction.

Source

ProductionEvidence: Medium50/100

ASAPP GenerativeAgent: airline customer service automation using Amazon Bedrock + Amazon Connect

Use case typeCustomer service automationUpdated Jun 13, 2026

ASAPP built GenerativeAgent for airlines to automate complex customer service across voice and digital channels. The platform uses Amazon Bedrock for model orchestration and Amazon Connect for voice initiation, with API integration to airline systems and human-in-the-loop routing.

Organization: A major U.S. airline
Industry: Logistics
Location: United States
Published: November 2025

Reported outcomes

+21%

First-contact resolution improvement vs baselineOther quantified impact

39 daysImplementation time91%First-contact resolution92%Customer satisfaction

Strategic outcomes

New product / capabilityAutomated complex airline customer serviceCustomer experience & trustImproved customer self-service adoptionSpeed & agilityEnabled rapid solution implementationCustomer experience & trustImproved customer satisfaction

Why do we believe this?Outcome claims, sources, and evidence checks

Normalized claim

Implementation time: 39 days

AWS BlogNov 7, 2025Blog postExplicit claimMedium evidence strength

Completed in 39 days

Normalized claim

First-contact resolution: 91%

AWS BlogNov 7, 2025Blog postExplicit claimMedium evidence strength

91% first-contact resolution rate

Normalized claim

First-contact resolution improvement vs baseline: 21% increase

AWS BlogNov 7, 2025Blog postExplicit claimMedium evidence strength

representing a 21% improvement from baseline

Normalized claim

Customer satisfaction: 92%

AWS BlogNov 7, 2025Blog postExplicit claimMedium evidence strength

92% customer satisfaction scores

Why do we believe this deployment?Customer identity, provider attribution, maturity, and source checks

Customer: A major U.S. airline
Provider: AWS
Maturity: Production
Linked source: AWS Blog

It connects to Amazon Connect for voice interactions and to airline operational data via APIs

Customer identity supportedSource describes one deploymentMaturity supported

Primary read

Use case focus

Showing 3 of 3

1Customer service automation
2Contact center AI
3Agent assist

Airlines faced rising and complex customer-service demand across voice and digital channels.
Legacy IVR and rule-based chat could not handle complex workflows such as flight changes and refunds.
Customer service teams had to navigate fragmented airline systems manually, slowing resolution and increasing costs.

ASAPP's GenerativeAgent orchestrates foundation models in Amazon Bedrock for safety validation, reasoning, tool calling, and output verification.
It connects to Amazon Connect for voice interactions and to airline operational data via APIs.
The solution maintains interaction history, redacts sensitive data, and routes cases between automation and human agents when needed.

Technologies

Amazon Bedrock Amazon Connect

The implementation was completed in 39 days.
The airline reached 91% first-contact resolution, a 21% improvement over baseline.
It also increased customer self-service adoption and achieved 92% customer satisfaction.

Architecture

GenerativeAgent uses Amazon Bedrock to orchestrate safety validation, reasoning, tool calling, RAG, and output verification across different foundation models. Voice interactions start through Amazon Connect, while an adaptation layer connects to airline operational systems via APIs and redacts sensitive data. Cases are routed between automated handling and human agents when needed.

Sources & evidence1

Evidence: Medium50/100Evidence strength