GenAI in Motion Redefining Automotive - Innovating with Customer Insights

In the era of explosive growth in GenAI, we are witnessing transformative changes across all industries. Corporations are finding their ways to transform into innovative powerhouses by enhancing employee and customer experiences. At NTT DATA, we are committed to continuous innovation to provide advanced and effective solutions for our clients. We collaborate with clients to uncover valuable GenAI applications, identifying employee and customer journey-specific scenarios and pain points, and empowering various departments to significantly enhance business efficiency and reduce costs.

The GenAI Empowered Voice of Customer (VOC) solution, developed by NTTDATA, can revolutionize how businesses interpret and respond to customer voice of multiple formats from multiple channels. This solution is a comprehensive framework that utilizes advanced AI technologies to gather, analyze, and act on customer feedback and sentiments, thereby enabling businesses to respond proactively to customer needs and market trends. NTT DATA offers a service-oriented VOC solution instead of a SaaS product, providing a more data-compliant, real-time, and scalable approach with omni-channel data integration, adaptive learning capabilities, and immediate actionable AI insights, compared to conventional VOC's limited, periodic, and structured data analysis.

Business Scenarios

This business scenario framework offers a comprehensive overview of how specific business scenarios across various phases can be empowered by the VOC solution. NTT DATA VOC solution empower all business units of the enterprises. It covers strategic planning, product development, manufacturing, marketing, sales, customer service, and relationship management. By integrating VOC into these processes, businesses can achieve greater efficiency, enhance customer experience, and improve decision-making capabilities.

Take product scenario empowerment to illustrate the closed-loop management:

Identify product related voice, reflect overall product volume trends, highlight product related trending topics, analyze channel product sales, analyze product competition and market dynamics, etc.

Step 1: Voice Monitoring & Anomaly Indicators
The VOC system monitors customer voice and identifies anomaly indicators in real time through managing comprehensive dashboard and a concrete indicator management mechanism. It tracks product issues and satisfaction.

Step 2: Issue Categorization & Task Distribution
Issues are identified and categorized, compare to find best and worst performers, pinpoint key problems, analyze customer feedback, visualize findings, and understand core experience issues. Tasks are then distributed accordingly if the issues should be resolved through a work order. Alternatively, alerts are systematically sent if the issues enter a monitoring period that requires attention. For instance, the quality comparison, identify quality level, quality trends, top issues, and data source structure, in order to analyze anomalies of models.

Step 3: Issue Reception → Issue Follow-up → Task Closure Request
The relevant team receives and follows up on the issues. Depending on the type of issue (e.g., in-car system issue, hardware recall initiative, or product design iteration), the team addresses the problem and tracks the progress. Once resolved, a task closure request is initiated.

Step 4: Satisfaction Tracking & Task Closure
Product performance improvement is assessed through indicator monitoring to evaluate the effectiveness of issue resolution. If performance improves to a certain level, it indicates that the issue resolution has significantly enhanced product performance and satisfaction, thereby providing evidence to close the task.

Take service scenario empowerment to illustrate the closed-loop management:

Observe data trends and voice sentiment trends, understand regional service conditions, analyze customer complaint handling, analyze positive and negative service experiences, Identify feedback on service quality, etc.

Step 1: Anomaly Indicators - Complaint Rate Increase
The VOC system identifies anomaly indicators and captures complaints in real time through managing comprehensive dashboard and a concrete indicator management mechanism. It tracks product issues and customer satisfaction.

Step 2: Issue Categorization & Task Distribution
Once anomalies are detected such as the complaints, they are categorized into specific complaint types. This categorization is crucial for visualizing the current state of service experience issues. Following this, tasks are distributed accordingly, either through work orders for immediate resolution or systematic alerts for issues that require a monitoring period.

Step 3: Task Acceptance and Processing
The relevant team receives and follows up on the issues, addresses the problems, and tracks the progress. Once resolved, a task closure request is initiated.

Step 4: Negative Complaints Reduced & Task Closure
The certain reduction in negative complaints and an increase in performance metrics indicate that the issue resolution has enhanced service satisfaction. This provides the evidence to close the task, ensuring a close-loop solution to the service.

The Marketing and Public Relations Management is a special scenario that not only exemplifies the closed-loop management, but also highlights the seamless integration with AIGC:

Identify trending discussion topics, analyze topic sentiment, identify high-frequency topic keywords, detect and address public relations issues, and track the effectiveness of issue-solving outcomes. Analyze overall activity performance, understand activity themes' impact, and trending feedback, etc.

Step 1: Anomaly Indicators - High Total Negative Voice Ratio
A significant increase of Total Negative Voice Ratio is identified, which serves as an anomaly indicator for potential issues affecting brand or product, highlighting the need to review the negative sentiment.

Step 2: Issue Categorization & Task Distribution
Issues are categorized based on the nature of the negative sentiments. Tasks are then distributed accordingly, either through work orders for immediate resolution or systematic alerts for issues that require a monitoring period. For instance, analyze the data sources where current issues are most concentrated, in order to implement targeted improvements.

Step 3: Task Acceptance and Processing
Tasks are accepted and proceed by the relevant team. During this process, AIGC is adeptly leveraged to create PR and marketing content that is meticulously tailored to align with the unique style of each specific channel, such as Facebook and TikTok. AI-generated content is utilized to engage with our audience across social media and PR platforms, addressing concerns, sharing updates, and restoring confidence in brand. This strategic approach enhances the appeal of the content and significantly improves channel distribution efficiency.

Step 4: Total Negative Voice Ratio Decrease & Task Closure
The certain reduction in negative sentiments indicates the appropriate actions. This provides the evidence to close the task, ensuring a close-loop solution to Marketing and PR.

Take quality scenario empowerment to illustrate:

Identify key quality and maintenance issues, understand the geographical distribution of issues, uncover competing products' quality feedback, observe channel trends in quality-related feedback, etc., to fulfill risk warning and management.

Take in-depth competitor analysis scenario empowerment to illustrate:

It could include both market competition analysis and product competition analysis. It contains indicators from various dimensions such as sales data, competitor highlights, equipment comparisons, top issues, performance metrics, and feedback analysis. It identifies key competitors and analyzes customer purchase intentions, reasons, and satisfaction levels. An in-depth analysis of selected models is conducted to identify competitive factors, including winning and losing elements.

AI Agent

At the heart of our VOC solution lies the AI agent technologies, including RAG, Prompt Engineering, Embedding, Knowledge Graph, and LangChain, which provide the cognitive framework and processing power that enable AI agents to efficiently analyze, understand, and applied effectively of customer voice.

The AI Agent streamlines the processing of customer voice by first transforming the structured information of documents or data into embedding vectors. This conversion allows the AI system to process and navigate through the content efficiently, identifying entities within the text. By embedding these into vectors, the AI system gains the ability to comprehend the deeper meanings and connections within the text. The Adapter links these embedding vectors to the tagging system, ensuring precise information clustering, while textual tokens are further analyzed to identify key information and concepts.

Starting with Data Filtering, the AI evaluates the source data to ascertain its validity. During the Tagging process, the AI identifies data that matches tags within the Tag Library through AI Selection. Concurrently, data that does not correspond to any existing tag is filtered, reviewed, and if appropriate, added to the Tag Library. This mechanism ensures a dynamic and precise tagging system that is constantly evolving and refining the Tag Library to enhance data categorization and retrieval. Content flagged by AI for negative sentiment undergoes further analysis and mining to identify underlying issues. These issues are then matched with the existing Common Question Library. If a match is found, the issue is merged with the existing issue. However, if no match is identified, the issue is automatically clustered by AI as a new problem, which ensures the Common Question Library is continually updated.

Large Language Model

The VOC solution leverages large language models (LLMs) and diverse AI services. NTTDATA algorithm services include data labeling, training, evaluation, sample management, and model fine-tuning, ensuring robust and accurate AI models. The algorithmic model encompasses advanced functionalities like emotion analysis, keyword extraction, semantic analysis, entity tagging, composite tagging, opinion and comment extraction, high-concurrent predictive task support and text classification to comprehensively understand customer voice. NTTDATA on-premise AI services offer cutting-edge capabilities, including open-source LLMs for semantic analysis, stable diffusion model for text-to-image conversions, multi-modal models for media processing, and embedding models for text vectorization. Additionally, we integrate third-party AI services from leading providers such as OpenAI, Google, and Meta, Kimi, Ernie Bot, enhancing our solution with their state-of-the-art technologies.

Data Source

In the VOC system, data sources are primarily divided into 1st Party Data and 3rd Party External Data. 1st Party Data encompasses information directly collected from various customer interaction channels, including call centers, e-commerce platforms, community feedback, and customer surveys. This data includes customer complaints, outbound surveys, online customer service interactions, in-store voice recordings, purchasing behavior, product reviews, comments, and in-car experience feedback. 3rd Party External Data consists of industry reports and social media insights. Industry reports provide vertical media data and purchased reports that offer broader market insights. Social media data, gathered from platforms such as TikTok, Redbook, and Facebook, along with search engine results, captures customer opinions from external sources. Through the application of ETL processes, this diverse data is extracted, refined, and transformed into a structured format that is conducive to analysis. The ETL methodology-short for Extract, Transform, Load-serves as the backbone of our data integration strategy, systematically pulling data from disparate sources, converting it into analyzable information, and subsequently warehousing it within the database. API is utilized as well for facilitating the extraction of this diverse data, and Web Scraping is also harnessed to extract information, especially suitable for sites that do not offer an API interface. Each data type is well transformed and integrated into a coherent format that facilitates comprehensive data analysis. This conversion and processing of multi-modal data ensure that diverse sources of information are uniformly prepared and analyzed, providing a holistic view of the customer insights.

In conclusion

The GenAI Empowered Voice of Customer (VOC) solution by NTTDATA represents a paradigm shift in the realm of customer experience improvement and business empowerment. It stands as an exceptional testament to the seamless fusion of advanced AI with the holistic understanding of customer-centric data. This solution encapsulates the essence of proactive business strategies, offering a 360-degree perspective on customer feedback and sentiment through its multi-modal, multi-channel data acquisition and transformation processes, achieving the closed-loop management and effective operation.

It not only captures the voice of the customer but also translates it into actionable strategies that are aligned with the dynamic needs of the market which is the core application capability of VOC.

The charm of the VOC solution also lies in the fact that its application extends beyond this, offering a multitude of scalable business applications. We truly recognize the potential of VOC solution to elevate businesses to new heights of customer satisfaction, market responsiveness and corporate operation. It is a beacon of innovation, guiding enterprises through the complex landscape of customer voices with clarity, precision, and an unwavering commitment to excellence.

Megan Wang

Megan Wang

Vise President of Digital Marketing, NTT DATA China

Contact

NTT DATA Group Corporation
Global Marketing & Communication Headquarters
Shinsuke Yoshinaga, Shota Yano
E-Mail: global-marketing@kits.nttdata.co.jp