A leading Indonesian telco transforms customer experience with Neo, OBZ's enterprise AI Coworker platform
How OneByZero deployed Neo, its enterprise AI Coworker platform built on Amazon Bedrock with Anthropic's Claude Sonnet, to power 50+ automated customer journeys, including agent-assist and voice, for one of Indonesia's largest telcos, serving approximately 95 million subscribers.
Powered by Amazon Bedrock with Anthropic's Claude Sonnet
7×
improvement in chatbot NPS, from an average of 4 to 30
92%
First Contact Resolution, up from 59%
34%
reduction in agent ticket handling costs through automation
Overview
About the customer
The customer is one of the largest integrated telecommunications providers in Indonesia, serving approximately 95 million subscribers across mobile, home broadband, and enterprise services. Operating in a highly competitive, multilingual market, the company manages large-scale digital and contact centre operations under multiple consumer and business brands. As part of its digital transformation strategy, the operator set out to modernise customer service and introduce AI-driven automation to improve experience, efficiency, and scalability.
The challenge
Scaling customer engagement in a multilingual, high-volume market
The operator's digital service channels were underperforming against traditional contact points on satisfaction and resolution. A rule-based chatbot delivered inconsistent quality and leaned heavily on human agents, creating four critical pressure points.
01
Inconsistent rule-based chatbot
The existing bot delivered inconsistent resolution quality, driving low satisfaction and heavy dependence on human agents.
02
Chatbot NPS averaging just 4
Against 40+ for other contact channels, the chatbot was actively eroding customer trust.
03
First Contact Resolution at 59%
Unresolved queries drove repeated interactions, high agent handovers, and rising operational costs.
04
Fragmented, weakly governed journeys
Journeys were split across chat, voice, and digital, with limited Bahasa Indonesia / English support and no enterprise-grade governance.
The solution
Neo, OBZ's enterprise AI Coworker platform
OBZ designed and deployed Neo, an enterprise AI agent platform built natively on AWS, covering 50+ customer journeys, from sharing information and resolving product queries to recording complaints and raising service tickets, extended with agent-assist tooling for human agents and a dedicated voice agent channel. The solution leverages Amazon Bedrock as its AI platform, with Anthropic's Claude Sonnet serving as the core generative AI model, selected for the quality, consistency, and multilingual strength of its responses. Neo is structured across three layers, each with a distinct responsibility.
This engagement marked the first production deployment of Neo's Studio layer. Insights from the rollout directly shaped the platform's subsequent development, making this client OBZ's founding reference for Studio at enterprise scale.
#Layer 01
Foundry
AI & data foundation, owned by platform teams. Shared, reusable resources made available to every agent.
- Integration with enterprise knowledge bases, APIs, and customer systems
- Centralised data pipelines for conversation history, feedback, and usage signals
- Retrieval-Augmented Generation (RAG) pipelines for grounded, accurate responses
- Anthropic's Claude Sonnet on Amazon Bedrock, with Amazon Lex, for model access and orchestration
#Layer 02
Studio
Low-code agent and journey builder where agents are built, tested, and published to production.
- Visual journey builder for chat, voice, and agent-assist workflows
- 50+ journeys spanning information sharing, query resolution, complaints, and service tickets
- Low-code configuration of intents, prompts, personas, and business rules
- Reusable templates via the Journey Bank, with built-in testing and preview, created specifically to accelerate future replication across telcos
#Layer 03
Command Centre
Operations, governance, and optimisation, with real-time visibility across all live agents.
- Real-time performance monitoring and SLA tracking
- Agent handover management and escalation controls
- Integrated feedback, sentiment, and quality scoring
- Usage analytics, compliance auditing, and guardrail enforcement
Rollout
A phased, governed rollout
Neo went live in production with an initial scope of 21 journeys across one brand and one channel. Two additional brands and four additional channels are now in flight for production deployment, expanding the platform's reach without compromising the governance built into the initial launch.
AWS services used
Built natively on AWS, end to end
Neo runs entirely on AWS-native services spanning model inference, orchestration, retrieval, data, security, and operations, giving the operator enterprise-grade governance for sensitive interactions.
Amazon Bedrock
Hosts Anthropic's Claude Sonnet, the core generative model, for conversational intelligence and streaming responses, with Bedrock Guardrails for governance.
Amazon Lex
Multilingual natural language understanding for Bahasa Indonesia and English.
Amazon EKS
Containerised orchestration, scheduling, evaluation, and observability with autoscaling.
Amazon OpenSearch
Vector-based semantic retrieval for Retrieval-Augmented Generation (RAG).
Amazon RDS
Highly available transactional storage for session state and orchestration workflows.
Amazon S3
System of record for knowledge assets, campaign data, and analytics outputs.
Amazon ElastiCache
Low-latency session and context caching.
AWS VPC Endpoints
Private connectivity to AWS services without public internet exposure.
IAM, KMS & Secrets Manager
Access control, encryption, and credential governance.
CloudWatch + CloudTrail
Operational monitoring and auditability.
AWS CodeBuild
CI/CD automation for controlled deployments.
Delivered by OneByZero, an AWS partner, with security, identity, and compliance controls built in for regulated, high-volume telco operations.
Delivery
Delivered by a cross-regional team, with zero delays, defects, or escalations
The engagement was delivered by teams spread across India, Singapore, and the Philippines, coordinated through daily standups, weekly reviews with the project sponsor, and onsite support during critical go-live phases. The result: zero delays, zero defects, and zero escalations across the entire delivery.
Outcome
A chatbot customers actually trust
7×
chatbot NPS, to 30 (from an average of 4)
92%
First Contact Resolution, from 59%
34%
lower agent ticket handling costs
The Neo deployment delivered measurable improvements across customer experience and operational efficiency, with a sharp reduction in human intervention and manual workflows across digital customer service operations.
- Agent handover rates declined by 36%, reducing human workload
- Average customer engagement runs at approximately 25 messages per session
- New journeys deployed up to 50% faster using the reusable Journey Bank
- Across all brands and channels, the chatbot now handles approximately 120,000 messages, 16,000 active users, and 30,000 tickets on a daily average, with 57 million total users served to date
- Bot traffic increased by 29%, indicating strong customer adoption
About OneByZero
OneByZero is a Frontier Systems Integrator specialising in AI Coworker design, build, and deployment for regulated enterprises across financial services, telecommunications, and retail. We operate across ASEAN, India, ANZ, and Japan, with a presence in the United States, combining domain expertise in regulated industries with deep AI engineering capability.
HQ Singapore · Tech HQ USA · Australia · India · Indonesia · Malaysia · Philippines · Thailand · Vietnam
