Workflows & Automation

How U Mobile used Amazon Bedrock AgentCore to build agentic telecommunications operations.

U Mobile built agentic helpdesk AiDaH on Amazon Bedrock AgentCore to resolve 150 support scenarios and cut ticket volumes by up to 84.5 percent.

How U Mobile used Amazon Bedrock AgentCore to build agentic telecommunications operations.

Key Takeaways

  • U Mobile deployed AiDaH, an AI-driven agentic helpdesk built on Amazon Bedrock AgentCore with AWS and Axrail, to automate telecommunications support across more than 150 operational scenarios.
  • AiDaH can investigate issues across multiple backend systems and propose remediation for human confirmation, going beyond traditional AI tools that mainly help agents draft replies.
  • U Mobile reports significant operational impact, with internal measurements showing trouble ticket volumes falling by up to 84.5 percent in key categories.
  • The platform uses a phased, “think-big, start-small” approach, expanding from pilot agents to a multi-agent system supporting six business divisions.
  • Its architecture connects Microsoft Teams, AgentCore orchestration, eight specialist agents, Model Context Protocol gateways, and governed knowledge bases to enable secure, cross-system workflows.

Meta description: U Mobile deploys AiDaH, an agentic helpdesk on Amazon Bedrock AgentCore, to resolve support tickets across 150 scenarios and cut ticket volumes.

How U Mobile used Amazon Bedrock AgentCore to build agentic telecommunications operations.

Tags: AiDaH, Amazon Bedrock AgentCore, U Mobile, agentic AI, operations automation

Featured image: an abstract, friendly illustration of connected service agents, cloud infrastructure, and a support chat interface, with no real people, logos, or text.

For U Mobile support staff and enterprise technology readers, AiDaH offers an AI-Driven Agentic Helpdesk that can investigate problems across multiple operational systems instead of only suggesting replies. U Mobile, Malaysia’s newest 5G network provider, has deployed the platform on Amazon Bedrock AgentCore with AWS and AWS Partner Axrail. U Mobile’s internal measurements show trouble ticket volumes falling by up to 84.5 percent in key categories.

From lean automation to operational agents

U Mobile has long kept telecommunications operations lean, using continuous automation to do more with fewer resources. When new artificial intelligence capabilities became available, the Information Systems Division looked beyond basic productivity gains. The larger opportunity was earlier in the support workflow: helping operations teams diagnose and resolve problems, not just compose better responses.

Traditional automation tools can help support agents draft replies. They do not navigate multiple systems, analyze results, and propose corrective action. Closing that gap requires an agentic architecture. AiDaH, short for AI-Driven Agentic Helpdesk, was built on Amazon Bedrock AgentCore, a platform for building, connecting, and optimizing agents at scale with any framework or model.

The team describes AiDaH not as a one-off technology initiative, but as the next step in a long-running effort to simplify operations and scale capability without adding complexity. Two platform capabilities shaped the choice. The managed agent runtime removed the need for U Mobile to operate separate agent infrastructure. Built-in memory carried conversational context across turns without a separate state store. Within a year of starting the journey, AiDaH handles over 150 documented operational scenarios across six business divisions.

A phased build across U Mobile’s information technology estate

U Mobile’s information technology environment reflects common telecommunications complexity: business support and customer relationship management systems, mobile number portability platforms, big data analytics, and infrastructure monitoring tools running across on-premises and cloud systems.

The Information Systems Division strategy rested on three pillars. Reach meant agents that coordinate across backend systems in real time. Depth meant multi-step root cause analysis using chain-of-thought reasoning. Community meant a secure multi-tenant framework with role-based access control and built-in guardrails available to every division.

The project began with a think-big, start-small principle. Conceptualization began in late 2025 with two pilot agents and Amazon Bedrock Knowledge Bases, a fully managed Retrieval Augmented Generation capability. A first-quarter 2026 pilot brought the internal information technology helpdesk online with 4 agents and 20 operational scenarios. Through 2026, the platform hardened and expanded, adding three-tier routing and reaching 6 agents by mid-year. AiDaH went live to 50 level-2 call center and branch users in third-quarter 2026 with 8 agents, and now covers more than 150 documented operational scenarios.

The difference between ordinary AI assistance and agentic work is visible in a common request. If a support agent handles “mobile data not working overseas,” a traditional assistant may help draft a response and suggest troubleshooting steps. The agent still has to manually check roaming status, verify eligibility, raise tickets, and send the final reply. AiDaH is designed to move the request through the relevant systems, ground the answer, and return a finding and proposed remediation for human confirmation.

Six layers connect Teams, agents, and backend systems

AiDaH runs on a layered architecture built on Amazon Bedrock AgentCore and deployed in the Asia Pacific (Singapore) Region. Staff use Microsoft Teams, a deliberate choice to meet users where they already work, avoiding a new application to learn or portal switching.

Layer one is the staff interface. Teams and a web portal served by Amazon CloudFront, protected by AWS WAF, reach an AWS Lambda backend and a REST API for programmatic triggers.

Layer two is orchestration. The routing layer on Amazon Bedrock AgentCore applies a pre-flight security check covering role-based access control, identity validation, and guardrails. It then uses a cost-optimized routing tier. Roughly 70 percent of requests resolve through deterministic keyword matching with no model invocation, 20 percent are classified by a lightweight model, and only 10 percent require the full orchestrator. The cheapest path that can confidently handle the request is the one that runs.

Layer three contains eight specialist agents: customer relationship management, customer problem management, number portability, order and subscriber identity module operations, infrastructure monitoring, big data analytics, cloud cost management, and finance operations. They communicate over the Agent2Agent protocol, allowing them to delegate subtasks and complete cross-system workflows that no single agent could handle alone.

Layer four is gateways and backend interfaces. AgentCore Gateway exposes U Mobile’s backends to agents through a unified Model Context Protocol interface. Behind each gateway sit AWS Lambda functions and Model Context Protocol servers that translate agent intent into calls against a specific system.

Layer five handles data and knowledge. Amazon Bedrock Knowledge Bases ground agent responses in governed content using an Amazon OpenSearch Serverless vector store. Content spans a shared enterprise knowledge base and a separate finance corpus. Amazon Bedrock AgentCore memory maintains conversational context across turns, and Amazon DynamoDB holds the six-tier role-based access control model and session metadata.

Layer six reaches U Mobile’s on-premises and cloud operational systems over AWS Transit Gateway and AWS Site-to-Site VPN, an Internet Protocol Security encrypted tunnel with no public internet exposure.

How one request travels through AiDaH

A port-in status question illustrates the flow. A user types the question into the AiDaH channel in Microsoft Teams. The Teams message reaches an AWS Lambda function, which forwards the request and the caller’s federated identity to the orchestrator.

The orchestrator resolves the caller against the six-tier role-based access control model in Amazon DynamoDB, validates the token issued by AgentCore Identity, and applies Amazon Bedrock Guardrails to the incoming prompt. A request that fails any check stops there.

The orchestrator then selects the cheapest tier that can handle the request. A recognized phrase such as “port-in status” can match deterministically and skip model invocation. An ambiguous phrase goes to the lightweight classifier. A novel or multi-part question goes to the full orchestrator.

The agent queries AgentCore Gateway for the tools it is entitled to use. Each gateway publishes backend capabilities as Model Context Protocol tools with typed inputs and outputs, so the agent selects a tool by name and schema rather than hand-coded integration logic. Amazon Cognito brokers the Open Authorization flow for the call.

The gateway’s Lambda function or Model Context Protocol server translates the tool call into a request against the target system. The path traverses AWS Transit Gateway and the AWS Site-to-Site VPN tunnel to reach the on-premises number portability platform. Sensitive fields are masked on the way back, before the response reaches the model.

If the answer spans domains, the agent delegates over the Agent2Agent protocol. A port-in query that turns out to be a billing block can hand off to the business support systems agent, and the results are merged into one narrative. The agent grounds its explanation against Amazon Bedrock Knowledge Bases, then returns the finding and proposed remediation to Teams for human confirmation. AgentCore Observability writes the full trace of prompts, tool calls, token counts, and latencies to Amazon CloudWatch, and AWS CloudTrail records the corresponding API activity.

The platform’s Amazon Bedrock AgentCore capabilities map to production roles in a clear way. Runtime hosts AiDaH and its sub-agents as managed containers that scale automatically, with no servers to provision. Memory maintains conversational context such as subscriber number, order IDs, and prior findings across turns. Identity issues short-lived access tokens so only authorized users can invoke agents. Gateway connects agents to U Mobile backends through a unified, secure Model Context Protocol interface. Observability streams every prompt, tool call, token count, and latency measurement to Amazon CloudWatch.

Onboarding a new operational domain adds a gateway and its Model Context Protocol tool definitions at the tool-discovery step. The earlier routing, security, interface, and response steps remain unchanged. The stable core keeps the marginal cost of each new domain low. The team notes that a spec-driven method was used to build the platform.

Security, auditability, and measured impact

Before scaling AiDaH to production, U Mobile engaged AWS to run a Well-Architected Security Review of the Landing Zone hosting the platform. The review validated the environment against AWS security best practices. Infrastructure protection came first through AWS Control Tower guardrails and AWS Organizations service control policies. The review also examined centralized identity federation and network security controls for ingress and egress. Consolidated security findings in AWS Security Hub and encryption standards across every architecture tier completed the assessment. The outcome was confidence that every AiDaH component operates on an independently validated foundation, plus a prioritized roadmap for continuous improvement.

Every prompt, tool invocation, and response is streamed to Amazon CloudWatch Logs through AgentCore Observability, and AWS CloudTrail records the underlying API activity. That audit trail matters because AiDaH is now written into U Mobile’s level-2 support standard operating procedures and is subject to operational audit. All compute runs in private Virtual Private Cloud subnets, on-premises connectivity is exclusively over encrypted virtual private network paths, and AWS Key Management Service manages encryption keys for data at rest and in transit.

The platform covers six domains. Customer management includes account details, billing history, payment status, active plans and contracts, subscriber identity module information, and service status. Order operations and logistics includes order and delivery status, pending orders, resource allocation, tracing workflows, delivery tracking, and shipment history. Customer problem management includes root-cause diagnosis for failed add-on purchases, plan changes, and top-up failures, as well as number portability port-in and port-out checks and service disruption diagnosis across short message service, calls, and data. Information technology infrastructure operations includes server and device status monitoring, IP and hostname lookups, and network device inventory. Business intelligence and analytics includes subscriber trend analysis, device take-up trends, revenue trending, anomaly detection, and individual subscriber analytics; the analytics agent converts plain-language questions into Structured Query Language queries, so no technical skills are required. Knowledge retrieval covers enterprise-wide retrieval with context-aware, grounded responses and citations back to source evidence.

For U Mobile, the result is an operations platform that extends automation from answering questions to resolving workflows. For readers evaluating agentic AI, the case shows how a lean team can add coordination across existing systems while keeping security, audit trails, and human confirmation in the path of every request.

Comments (0)

  1. No comments yet. Be the first to share what worked for you.

Leave a comment

Comments are reviewed before they appear. Your email address is not published.