A global automobile manufacturer operating across multiple markets manages millions of unstructured customer interactions — from surveys and social media to contact centre logs. Seeking to transform its customer experience strategy, the company partnered with Artefact to build a Holistic Voice of the Customer (VoC) platform powered by Generative AI and AWS, replacing slow, manual analysis with a scalable, insight-driven system that directly improved sales conversion and customer retention.

The challenge: manual VoC analysis blocking customer insight.

The company’s existing process for analysing customer feedback relied entirely on human analysts, who manually classified verbatims into a fixed, static taxonomy. The result was generic, low-resolution insights that failed to capture the nuance needed to drive meaningful business action across global markets.

  • Manual classification bottleneck: Human analysts classified customer verbatims one by one, producing vague topic definitions such as ‘Cleanliness of the Car’ rather than granular distinctions like ‘Vehicle Cleaning Issues’ vs. ‘Valet Service Issues’ — slowing insight generation and limiting analytical precision.
  • Blind spots on emerging topics: The static manual taxonomy could not surface previously unseen topics — such as the impact of climate change on brand perception — creating intelligence gaps that prevented the company from acting on emerging customer sentiment.
  • Inability to scale across markets: With massive volumes of unstructured feedback flowing from Sales, After-Sales, Google Search, and Contact Centres across multiple geographies, manual analysis was fundamentally unscalable and inconsistent across markets.

The solution: a Holistic VoC platform with LLMs and AWS DataZone.

Artefact architected and delivered a Holistic VoC platform combining unsupervised machine learning with Large Language Models (LLMs) on AWS — automatically discovering, classifying, and generating insights from unstructured customer feedback at global scale. The solution was productionised for maximum efficiency, leveraging a governed Hub-and-Spoke architecture managed through AWS DataZone.

  • Unsupervised learning for topic discovery: Applied embedding models and clustering algorithms to process raw customer verbatims, automatically identifying high-density topic clusters without prior knowledge or human bias — enabling discovery of topics the manual taxonomy had never anticipated.
  • Few-Shot LLM classification pipeline: Deployed a Few-Shot learning pipeline using Foundation Models to classify new customer reviews into the established taxonomy (Topic, Subtopic, Sentiment) at scale, delivering near real-time insight generation in production.
  • AWS DataZone for global governance: Leveraged AWS DataZone as the central governance layer to securely distribute AI models from the Global hub to Local Market environments, treating models as governed data products and ensuring quality validation before each regional deployment.
  • Productionisation and cost optimisation: Led a dedicated productionisation phase to refactor the initial PoC architecture, achieving an 83% reduction in prediction time and a 90% reduction in monthly execution costs — enabling a cost-effective, enterprise-grade production deployment.

The results: measurable lift in conversion, retention, and efficiency.

The Holistic VoC platform delivered quantified, statistically validated impact across conversion, retention, and operational cost — demonstrating the direct commercial value of Generative AI deployed responsibly at scale.

  • +5.9% increase from base conversion rate: Validated through rigorous A/B testing at 98% statistical significance, with high-value segments such as ‘Premium and Owned Residences’ achieving conversion lifts of up to +11.6% — directly attributable to AI-driven Next Best Action recommendations.
  • 90% reduction in monthly operating costs: The productionised solution reduced monthly execution costs by 90% compared to the initial MVP/PoC phase, establishing a highly cost-efficient production baseline as the platform scales to additional markets.
  • ~83% reduction in prediction time: Processing time was reduced by five-sixths, enabling the company’s teams to access customer insights significantly faster and respond to emerging issues in near real-time across global markets.

By replacing manual analysis with a governed, AI-powered VoC platform on AWS, Artefact enabled a leading global automobile manufacturer to transform customer feedback into a strategic asset — delivering measurable ROI, scalable global intelligence, and a replicable architecture ready to expand to new markets as the company’s customer experience ambitions continue to grow.