At a high level, the EdgeTier AI system works by ingesting multi-channel conversation data from integrations with customer support platforms. It then processes this data using advanced NLP models to analyse and extract insights such as customer sentiment, key topics of discussion, and anomalies (unusual activity in the contact centre). These insights are used within the platform to generate actionable information and conversation summaries, helping businesses to enhance their customer service and operational efficiency.
If you’re interested in how EdgeTier is compliant with the EU Artificial Intelligent Act (2024), you can read our explainer here.
Anomaly Detection
EdgeTier's AI system continuously monitors customer service channels, identifying unusual spikes in specific tags or deviations from typical patterns based on data from the previous seven days. This feature aids in the early detection of issues such as product defects, technical problems, or policy concerns, enabling proactive management. The system sends real-time alerts to the relevant personnel when anomalies are detected, facilitating swift action.
Sentiment and Emotion Analysis
EdgeTier's proprietary AI models analyze customer messages across all languages, classifying them into categories such as Praise, Gratitude, or Frustration. These models are tailored to specific industries, ensuring accuracy and relevance, and the emotions that EdgeTier focus on are those that are commonly associated with positive and negative customer experience scores.
Semantic Similarity Search
The system supports multi-lingual semantic search, enabling users to find interactions with similar meanings to a loosely defined phrase, even if the exact wording varies.
Phrase Tagging
Custom machine learning models tag interactions based on a predefined set of keywords and phrases. These tags can be applied across all languages, helping to identify relevant topics as defined by the user. Tagged interactions can trigger immediate alerts, be integrated into the customer's existing system, or be reviewed later for detailed analysis. Where custom phrase tags are required, the inputs for each phrase tag setup is created individually for each customer, and are only used to output phrase tags for that customer.
GenAI Interaction Summaries
Using Claude AI and custom prompts, our system generates concise summaries of customer interactions, highlighting key aspects such as the main reasons for contact, sentiments, and resolutions. This feature allows users to quickly understand the essence of customer interactions without reading entire transcripts. Interaction summaries are stored in EdgeTier's database, presented in the EdgeTier UI, and can be integrated into customer support systems.
GenAI Auto-Responses
EdgeTier can create a database of frequently-asked questions and their approved answers, specific to each customer’s industry, agent guidelines, and business processes. This database then trains AI models to generate accurate, on-brand responses to customer emails or chat responses. The responses can be sent automatically to the bulk of customer interactions, or presented as options to agents, depending on customer preference.
Experience Scores
The EdgeTier system processes data from sentiment analysis and operational metrics to compute an overall Experience Score for every conversation we ingest. This score provides a quantitative measure of customer satisfaction and service quality even when the customer hasn’t manually replied to NPS or CSAT surveys.
Automated Contact Reason Labelling
The EdgeTier system can use generative AI systems to automatically determine the primary contact reason from a chat, call, or email transcript from a selection of reasons specified by the customer. This function reduces the administrative overhead on customer service agents and improves the consistency, accuracy, and coverage of contact reasons for reporting and insights.
Spotlight Summaries
The Spotlight Summaries feature leverages the Interaction Summaries feature by generating an instant AI-written overview of multiple customer conversations. It aims to highlight common issues and trends within a set of interactions. Users can apply filters, then click ‘Spotlight’ to see a high-level summary without reviewing each interaction individually.