Accelerating Inquiry Classification and Operational Efficiency with IBM Watson Natural Language Understanding
A specialized software provider in Japan wanted to reduce the manual workload involved in categorizing thousands of customer inquiries across multiple channels. Their existing process relied on human review, which slowed down response times, increased operational costs, and made it difficult to identify emerging trends.
Partnering with Nexright, the organization implemented IBM Watson Natural Language Understanding (NLU) to automate inquiry classification, extract key conversation insights, and strengthen decision-making across their support teams. The AI-powered model significantly improved prediction accuracy, reduced manual effort, and provided new visibility into customer sentiment and inquiry themes.
Business challenge
The company handled a high volume of inquiries daily—ranging from product questions to service support—for small and midsized business clients. However, the categorization process was entirely manual, leading to:
Key Challenges:
- Large fluctuations in inquiry volume, making it difficult to maintain consistent service levels.
- Manual classification bottlenecks, resulting in slower customer response times.
- Limited visibility into inquiry trends, impacting product planning and customer experience improvement.
- Increasing labor costs, with employee time heavily consumed by repetitive classification tasks.
The organization needed a scalable, accurate, and automated way to classify inquiries and support continuous operational improvement.
Solution
Nexright helped the client evaluate multiple AI options and determine the best-fit approach using IBM Watson Natural Language Understanding.
The solution used Watson NLU to analyze sentences, detect relevant features, extract keywords, and accurately classify messages into the correct categories—reducing the need for human review.
A proof of concept (PoC) demonstrated strong accuracy and suitability for real-world usage. Using Watson NLU combined with training on the client’s historical inquiry data, the team built a customized language model capable of understanding domain-specific terminology, slang, and context.
Once implemented, the automated classification workflow delivered:
- Significant reduction in manual review workload
- Faster identification of topic trends
- Improved customer service efficiency
- A strong foundation for future AI expansion (e.g., generative AI assistants)
Solution components
- IBM Watson Natural Language Understanding
- IBM watsonx.ai (for training, fine-tuning, and model management)
Automated Inquiry Classification
Watson NLU categorizes incoming messages using advanced linguistic analysis, correctly identifying query intent even in short or ambiguous messages.
Custom Domain Language Model
Nexright and the client co-created a model trained on the organization’s own historical data, ensuring high accuracy for industry-specific terms.
Keyword Extraction & Trend Analysis
Key terms and patterns are automatically surfaced to help the team identify emerging issues, customer needs, and product opportunities.
Result
- >51% improvement in prediction accuracy after training the Watson NLU model with custom data
- Significant reduction in manual classification time, enabling support teams to focus on higher-value tasks
- Faster customer response times, leading to improved satisfaction
- New insights into inquiry patterns, driving proactive service improvements
- Lower operational costs, as repetitive work shifted away from human agents
The solution created a measurable improvement in business efficiency while enabling a sustainable, AI-powered support model.
With Watson, we can automatically categorize inquiries with far greater accuracy and speed. The AI-powered workflow allows us to focus on customer needs instead of manual processing. Nexright helped us quickly identify the most effective model and deploy it into production seamlessly.
— Lead IT Manager, Japanese Software Provider