Customer Sentiment Tracking with AI

OUR SOLUTIONS AT WORK Machine Learning Ops | Generative AI | Natural Language Processing

Customer Sentiment Tracking

The client is a technology leader connecting home builders with prospective buyers for over two decades, offering services such as lead generation, digital marketing, and 2D/3D visual content. They take pride in pioneering digital technologies to drive their industry forward.

Objectives

The Customer Experience (CX) team aimed to address customer inquiries in a timely manner, but some messages were inadvertently overlooked or received delayed responses. This was occasionally leading to negative experiences and potentially undermined customer satisfaction.

To address this problem and ensure timely responses, the client sought an alert system integrated into their Basecamp installation that could:

Intelligently identify recent customer comments that would benefit from an priority/escalated response (per custom criteria).

AND

Track comments which haven’t been answered within 7 days.

Solutions

EX Squared’s AI solution surfaced negative feedback and unanswered comments needing urgent responses.

Using Basecamp APIs, conversational data between customers and the CX team was collected and cleaned.
Conversation data was annotated based on project stakeholders’ criteria, to label each as ‘Flagged’ or ‘Not Flagged’ for priority response.
BERT-based AI models were trained to categorize and flag client comments. With stakeholders’ feedback, our system achieved 95% accuracy flagging messages needing a priority response, allowing for a 2-3% margin of error.
Considering clients’ data sensitivity, we deployed the solution on-premises with Docker, generating a daily report highlighting flagged comments.

Results

EX Squared’s solution helped stakeholders monitor CX account managers’ responsiveness, resulting in faster interventions, improved customer experience and reduced client churn. Key result: 48% reduction in negative client comments (2021) compared to the previous year, followed by a subsequent 35% reduction (2022). Based on these strong results, additional features were implemented to track surveys and replies.

Talk with us

Let us know how we can
help shape your future.