Automated Technical Support with AI Agent
Industria
Telecommunications
Escenario
AI Agent
Cliente
Confidential
The client needed a solution to automate technical support for its Automated Recharge System application, used by thousands of customers for service top-ups and payments. The manual process created high operational workload and slow response times, especially for Level 2 and Level 3 support teams.
Solution
We implemented an AI-powered virtual agent that acts as an automated triage system. It processes emails forwarded by agents and transforms them into structured, clear, and actionable support tickets for the technical team.
Key Features
- Multimodal analysis (text + images) of incoming emails
- Automatic extraction of error type, affected phone number, screenshots, logs, and customer data
- Smart classification into three categories: phone services, electric services, or Electronic Vehicle charging
- Verification of required data: customer ID, service number, email address
- Context validation to confirm if the case is truly related to Automated Recharge System.
Technical Integrations
Direct Integration with Automated Recharge System Platform
- Retrieves customer transaction history (last 10)
- Verifies registered cards and blacklist status
- Confirms whether the affected service is among the customer’s favorites
- Detects common errors such as status 60 (bank rejection)
Integration with Jira Service Management (Atlassian)
- Automatically creates, updates, and closes tickets
- Enhances traceability and collaboration between support levels
Integration with Jira Service Management (Atlassian)
Response Automation
The agent compares new cases with a historical incident database to suggest automatic solutions for frequent issues such as:
- Transaction rejections
- Blacklisted or inactive cards
- Services not marked as favorites
- Incomplete customer information
In future versions, the agent will automatically respond to or close unrelated SAR
The agent compares new cases with a historical incident database to suggest automatic solutions for frequent issues such as:
- Transaction rejections
- Blacklisted or inactive cards
- Services not marked as favorites
- Incomplete customer information
In future versions, the agent will automatically respond to or close unrelated SAR
Technology Stack:
- Language: Python
- AI Models: Gemini 2.5 Flash (multimodal)
- Frameworks: Langraph, Langchain
- Infrastructure: AWS Lambda, AWS Bedrock, AWS S3, AWS Cloudwatch, AWS CloudFormation.
- Support platform: Jira Service Management
