>90% agent evaluation accuracy
300% expansion in language support for voice and text
$100K in operational savings within weeks
Building a Better Audience Experience through Smart Contact Center Support
For an industry giant operating across markets, time zones and languages, contact center support plays a critical role in keeping operations running without interruption. As demand grew, so did expectations. Think: consistent quality, faster resolutions and better visibility into agent performance. But limited evaluation coverage, manual processes and inconsistent scoring made it difficult to scale efficiently while trusting performance insights. They realized the solution was not more automation. It was smarter automation. Together, we modernized their contact center on Amazon Connect Customer using AI-driven evaluations, Agent Assist and self-service capabilities.
Rewriting the Script for AI Accuracy
Demanding Precision
AI-driven quality evaluations across voice and chat channels were already deployed within our customer environment, an important step toward increasing capacity and improving consistency. When automated scoring accuracy plateaued at approximately 66.8%, however, it limited confidence in the data and reduced its effectiveness for coaching, performance management and operational decision-making. Scaling quality at speed remained out of reach without reliable insights.
They took a closer look at how evaluations were designed and applied, and their analysis surfaced several areas for optimization:
- Criteria grouped multiple behaviors into single questions, creating ambiguity for AI interpretation
- Instructions designed for human reviewers lacked the specificity required for consistent automated scoring
- Transcription gaps reduced the system’s ability to identify key terms and interactions
- The same AI approach was applied across all scenarios, even where simpler methods would be more effective
- Manual review processes limited coverage and slowed access to insights
With a clear understanding of the opportunity, the focus shifted to refining evaluation precision, expanding coverage and strengthening the role of AI in enabling more effective, data-driven support.
Reimagining Contact Center Support on Amazon Connect Customer
Lights, Camera, Scalability
We transformed our customer’s Amazon Connect Customer environment through Gen AI-driven Agent Assist automation, AI-powered agent evaluations and self-service capabilities. We redesigned the evaluation framework from first principles, tailoring each workflow to the method that would work best rather than forcing every question through the same model. This approach standardized processes, enhanced the reliability of AI-assisted insights and streamlined workflows to ensure efficiency and scalability.
Evaluation Accuracy: A Structured Optimization Approach
TGS applied a disciplined methodology to eliminate the root causes of low scoring fidelity:
- Atomic question design: Replaced complex, multi-behavior questions with focused, single-behavior questions to eliminate ambiguity and strengthen diagnostic precision.
- Method alignment: Mapped each question to the optimal automation method—combining generative AI, Contact Lens categories, and contact metrics.
- Precision instruction engineering: Developed explicit decision logic and structured prompts to guide AI interpretation with no ambiguity.
- Custom vocabulary configuration: Added domain-specific terminology and variations to Amazon Connect to optimize transcription fidelity.
- Iterative calibration: Implemented phased testing against human evaluators with continuous monitoring to lock in accuracy gains.
Agent Enablement and Self-Service Automation
The expanded solution gave agents the tools to resolve issues faster and supervisors the data to coach smarter:
- AI-assisted evaluations: A redesigned centralized framework combining deterministic scoring and GenAI-driven contextual questions per interaction.
- Amazon Connect Customer AI Agents: Real-time AI guidance integrated with ServiceNow and AI-powered workflows, optimized for agent productivity and accelerated issue resolution.
- Self-service automation: New end-to-end voice and chat experiences using customer APIs, reducing reliance on a live agent.
- Guided experiences: Multiple structured flows built using native Amazon Connect Customer capabilities.
- Unified workflows: ServiceNow integration and single sign-on (SSO) for streamlined agent experience.
- Expanded multilingual: 300% increase in language coverage across voice and chat interactions.
agent evaluation accuracy
expansion in language support for voice and text
operational savings within weeks
The Power Behind the Performance: An AWS and Amazon Connect Partner
Expertise That Turned Vision Into Results
The engagement delivered measurable improvements in service quality, operational efficiency and scalability by matching each business challenge to the right automation approach and continuously refining performance based on real-world results.
As a trusted AWS Premier Tier Services Partner with the AWS AI Services Competency, TGS designs and implements secure, cost-effective Gen AI and agentic AI solutions that deliver measurable business outcomes across cloud initiatives. As an Amazon Connect Service Delivery Partner, we specialize in helping organizations elevate customer experience journeys and business outcomes through Amazon Connect Customer.
That expertise, combined with a highly collaborative working model, gave our customer confidence at every step, accelerating progress while ensuring the solution was aligned to their goals from day one.
Interested in a similar solution?
Real-World Results
Evaluation accuracy on previously low-performing questions jumped from approximately 66.8% to 100% during testing; eliminating AI hallucinations entirely and aligning automated scoring with expert human judgment across every criterion.
But the transformation extended far beyond evaluation accuracy. The modernization effort delivered enhanced service quality, operational efficiency and scalability across the Amazon Connect Customer environment, creating value at every layer of the operation:
- Full coverage: Automated assessment expanded from limited manual reviews to all qualifying interactions, providing leadership with the comprehensive, data-driven visibility they needed.
- Sharper agent coaching: Atomic, granular scoring enabled more precise development plans and targeted feedback.
- Faster resolution times: Agent Assist and structured knowledge access reduced friction during live interactions.
- Expanded automation capabilities: Developed optimized end-to-end Agent Assist and self-service experiences that provided more scalability and reduced reliance on manual processes.
- Greater operational efficiency: Automation reduced manual QA effort, freeing supervisors to focus on high-value coaching.
- Increased agent engagement: Self-service access to performance data empowered agents to take greater ownership of their growth and performance.
The AI-driven evaluation capability quickly earned recognition as a standout success, delivering approximately $100K in productivity savings shortly after launching and setting the foundation for continued scale.
