PoC delivered in approximately 4 months using AQ-AI Agent.
End-to-end support from public web infrastructure setup through to trial launch.
NTT Business Solutions Corporation (hereafter "NTT BS") needed to automate and streamline their IT helpdesk operations — which handle a large volume of daily inquiries — and sought support in both building a generative AI-powered chatbot PoC (Proof of Concept*¹) environment and consulting on operational and cost optimization. AsiaQuest leveraged its proprietary AI agent template "AQ-AI Agent" to deliver a fully functional PoC environment in approximately four months, providing end-to-end support across AWS environment setup, AI implementation, security configuration, and technical knowledge transfer to the client.
*¹ PoC (Proof of Concept): A prototype and validation phase conducted to verify the feasibility of a new technology or concept.
Background & Purpose
Guiding users toward self-resolution of routine inquiries and expanding customer touchpoints (improving convenience)
IT helpdesk teams frequently face a significant burden responding to repetitive inquiries such as login procedures, forgotten passwords, and step-by-step operation guides. Many organizations have also struggled with the challenge of transitioning to support channels that better meet the needs of today's users, who lead increasingly diverse lifestyles and prefer to resolve issues quickly on their own.
To address these challenges, NTT BS — which provides a wide range of IT services — decided to implement a PoC to automate inquiry handling via an AI chatbot, with the dual goal of improving user convenience and reducing operational load. They selected AsiaQuest as their support partner for this project.
The key factors behind choosing AsiaQuest were the ability to propose rapid AI adoption leveraging the proprietary "AQ-AI Agent" template, and the speed at which that proposal was delivered.
Role
AI chatbot built rapidly using Amazon Bedrock and AWS CDK — end-to-end support from PoC operation through to in-house capability development
A lean, four-person team — PM, Lead SE, Infrastructure SE/PG, and Application PG — provided continuous support from requirements definition through operational readiness. The PM oversaw progress management, cost optimization proposals, and the handover plan, while each engineer handled their respective areas of design, implementation, and data analysis. This structure was designed to achieve both speed and quality within a tight timeframe.
Communication was built around two scheduled meetings per week. Rather than operating as a simple "client and vendor" relationship, AsiaQuest worked closely with NTT BS's infrastructure and operations teams as a true technology partner — engaging in ongoing discussions around cost optimization and in-house capability building, and supporting the client all the way through to establishing a self-sufficient operational structure.
The project was executed agilely across four phases, without heavy upfront requirements documentation.
- ①Architecture Design
- To handle sensitive data — including users' personal information — securely, the foundational system architecture was rapidly designed to meet NTT BS's requirements, along with non-functional requirements capable of supporting production operations.
- ②Cloud Environment Setup & AI Implementation
- Built on the "AQ-AI Agent" foundation, the cloud environment was automatically provisioned using IaC via AWS CDK. A RAG-based AI chatbot using Amazon Bedrock was implemented in a short timeframe, with all additional services also codified and integrated.
- ③Quality & Operational Readiness for Public Deployment
- In the PoC phase, priority was placed on reaching a deployable state rather than over-engineering. Infrastructure and application improvements were carried out in parallel, working hand-in-hand with the client to advance toward an operationally viable state within a short timeframe.
- ④ Knowledge Transfer & Foundation for Continuous Improvement
- To enable the client to move toward independent operation, practical hands-on sessions were provided covering topics such as AWS CDK deployment demos and secure login procedures.
System Overview
A seamless user experience enabling self-resolution at any hour
When users encounter a question, they simply type it into the on-screen chatbot, and the AI instantly responds in natural Japanese based on the organization's own knowledge — including FAQs and other resources. Users can resolve issues on their own even late at night or early in the morning, and proceed with tasks smoothly without needing to go through an operator.
- Login & Authentication Screen
- Chat Input Screen
- Admin Panel — FAQ Management & Chat History Screen
Technical Overview
- AI chatbot implementation (Amazon Bedrock + RAG configuration *²)
- Secure AWS infrastructure setup (IaC *³ / fully managed)
- Quality and operational readiness for public-facing deployment
- Knowledge transfer and foundation for continuous improvement
Designed for public-facing operation, the system was fully prepared for production use, including security hardening such as multi-factor authentication, audit logs, and threat detection. Full codification of the entire configuration also provides a foundation for rapid and accurate environment replication whenever specifications change in the future.
*² RAG: A method of having the AI reference organization-specific data — such as FAQs — to generate more accurate responses.
*³ IaC: A method of describing infrastructure configuration as code for automated provisioning and management.
Additional Proposals
Cost Optimization Roadmap：Multiple cost reduction proposals presented based on independently extracted and analyzed usage data, with minimal impact on business operations.
Data-Driven Operational Improvement：Ongoing improvement measures for AI response accuracy proposed through chat log analysis.
In-House Capability Support：Hands-on sessions for automated AWS deployment and secure login procedures to help the client build a self-sufficient operational structure.
Brand Experience
PoC launched across multiple sites — foundation laid for full-scale rollout
Under a tight schedule, the PoC was launched across multiple locations. A framework was established to collect real usage data — including which times of day and navigation paths users access the chatbot, and which FAQs are missing. A key achievement of this project was building a foundation that enables data-driven validation of AI accuracy and cost optimization.
Additionally, IaC implementation via AWS CDK and knowledge transfer helped the client establish the groundwork for independently building and operating environments going forward.
Outlook for the future
Continuous improvement driven by data
Based on usage data (resolution rates, AI response trends, etc.) and on-site feedback, the team is exploring ongoing improvements to response quality and operational optimization.
Expected Business Impact
By contributing to more efficient IT helpdesk inquiry handling and providing a convenient, modern support channel, the solution improves the user experience (UX) and contributes to higher service satisfaction.
Testimonial
Feedback From The Customer
Our core services include IT infrastructure monitoring and operations as well as IT helpdesk support, and we work every day to maintain stable operations and improve customer satisfaction. In this context, we explored the introduction of an AI chatbot as a new channel to further enhance inquiry convenience and expand customer touchpoints.
This PoC required us to simultaneously advance design, implementation, and operational readiness in a short period of time. Under these circumstances, we issued an RFP to multiple companies, compared and evaluated their proposals, and ultimately selected AsiaQuest based on requirements fit, feasibility, and scalability. The deciding factor was their quick grasp of our requirements and the forward-looking proposal they delivered that was well matched to what we needed.
The "AQ-AI Agent" accelerated the PoC launch, and the agile development style allowed us to cycle through implementation and review in short iterations — giving us a real sense of steady progress even within a tight timeline.
From a project management standpoint, timely communication alongside the regular weekly meetings allowed us to share status updates and issues as they arose. With many stakeholders involved, a WBS-oriented approach kept the entire project moving forward. For steps that required longer lead times, aligning on those in advance and building them into the schedule allowed us to reach all major milestones without delay — which we consider a significant achievement.
Given that this was intended for a public-facing web environment, ensuring the system was operationally ready from a monitoring, logging, and auditing perspective was equally important. Because the proposal was designed with operations in mind from the outset, the system was built to integrate immediately with our monitoring and operations infrastructure — making it a major value-add that the solution was viable as a complete operational system from the PoC stage onward.
Establishing a clear escalation path for inquiries the AI cannot resolve to human operators — and getting on-site operations up and running quickly — was also an area where our own strengths came through. Confirming in the PoC that the AI chatbot and human operators could be combined into a workable operational model was a meaningful step forward toward our next phase of planning.
Since our AWS builds had previously been primarily GUI-based, gaining exposure to IaC principles through AsiaQuest's support and transitioning to a more reproducible and easily modifiable structure was also a key outcome.
As part of the knowledge transfer, having everything set up so that we can continue testing, validating, and improving on our own after the PoC concluded is enormously valuable. It has given us the foundation to keep moving forward — making incremental UI adjustments, updating and tuning the RAG knowledge base, and carrying out other improvements as ongoing next steps.
Going forward, we plan to use the usage data and on-site feedback gathered during the PoC as a starting point for continuously improving response quality and operational efficiency, with the goal of further enhancing customer satisfaction.
Nobuhiro Okamoto, Section Manager
Managed Operation Center (Osaka)
Customer Success Division, Managed Services Department
NTT Business Solutions Corporation