AI in customer service: A case study by ONTEC AI and Sunny Cars
Customer service teams are increasingly facing intense challenges: rising customer expectations, growing complexity of inquiries, and the need to balance efficiency with quality. By implementing modern AI solutions, companies can not only optimize their service processes but also elevate the customer experience to a new level. This case study demonstrates how Sunny Cars optimized their operations for both employees and customers using AI in customer service.
About Sunny Cars
Sunny Cars is one of the leading providers of car rental services, with sales focuses in Germany, Switzerland, Austria, Italy, the Netherlands, Belgium, Luxembourg, and France.
The provider does not maintain its own fleet of rental cars; instead, it collaborates with a variety of car rental agencies in over 120 countries worldwide to offer customers a wide selection of rental stations and vehicles.
The company offers an all-inclusive car rental package with every booking, always including all necessary insurances and services to make the car rental process as transparent and straightforward as possible for customers.
For Sunny Cars, customer proximity and excellent customer service are of central importance.
The company places great value on making the car rental process as simple as possible for its distribution partners and end users. Therefore, this project focused precisely on the processes that directly contribute to the customer experience.
The challenge
The car rental market has been growing for many years, and so have customer inquiries at Sunny Cars.
Sunny Cars was therefore confronted with the challenge of efficiently managing the growing volume of customer inquiries while maintaining high quality and service excellence.
The inquiries covered a wide range of concerns, from forgotten items in rental cars to damage reports for reimbursement of the deductible, which Sunny Cars promises as a service, or customers who were not fully satisfied. The complaint rate at Sunny Cars is less than 0.8% of the total volume. However, many of the inquiries also fall into the category of “white noise” and need to be less actively processed, but rather tidied up.
The highly qualified Customer Care team at Sunny Cars had to read, manually classify, and process a large number of emails daily, which was extremely time-consuming.
New, equally qualified employees were not easy to find in times of labor and skilled worker shortages.
To handle these customer inquiries more efficiently and increase customer satisfaction, Sunny Cars decided to implement AI in customer service to assist with some of the daily tasks.
The ONTEC AI team was selected particularly because of its expertise in handling complex and sensitive data, LLM-agnosticism, full GDPR compliance, and the expert team that accompanied the process.
The solution
In collaboration with the ONTEC AI team, several solutions were implemented to support the Customer Care team:
Automated classification/categorization of inquiries
One of the first challenges was assigning inquiries to specific categories, known as ‘Claim Types.’
There were about ten different categories, such as ‘Credit Card,’ ‘Damage,’ ‘Feedback.’
This was previously done manually and took time. Therefore, this email classification was to be solved with AI support.
Based on historical data, various approaches were tested to automate this process:
- Classification inquiries to an existing LLM: Achieved a classification accuracy of about 20%.
- Self-trained models: Achieved an accuracy of up to 80%.
- Fine-tuning an existing model: Ultimately, an embedding model was used as a precursor, on which the classifier was built. This model could directly assign the correct category and achieved a practical accuracy of over 90%.
ONTEC AI helps customer service teams get their email inbox under control. The tool assists with categorizing, prioritizing, and even responding to emails. This saves the teams a lot of time and creates space for the truly important tasks.
Provision of summaries of inquiries
Another challenge was summarizing the incoming – often very detailed and extensively formulated – inquiries.
Previously, the Customer Care team manually wrote summaries to quickly recall the nature of the inquiry during later processing (e.g., after a response from the counterpart).
Automating the summaries was intended to speed up the processing by helping Customer Care employees quickly recall the nature of the inquiry at a glance.
Supported evaluation and prioritization through sentiment analysis
Different inquiries were previously manually prioritized based on various criteria.
To optimize the prioritization of inquiries, sentiment analysis was implemented.
This enables the automated recognition of the sentiment of customer inquiries (positive, neutral, negative, very negative).
This sentiment analysis supports the prioritization of concerns and allows for proactive responses to particularly urgent inquiries.
Future developments
In September 2024, the first parts of the project had already been implemented. In particular, the following solutions emerged from the initial needs analysis and are currently still in the implementation phase.
Claim type keyword classifier
Another project goal is to train a keyword-based classifier that recognizes specific case details such as ‘Tire Damage,’ ‘Engine Damage,’ etc.
These case details should be used to conduct detailed analyses and, for example, identify areas for improvement with partners and specific locations.
Response draft generator
Existing information should be fed into prompts to automatically generate response drafts.
These serve as a complete basis but can also be manually adjusted for the respective customer.
Voucher/refund classifier
This automatically provides a recommendation on whether a customer should receive a voucher, a refund, both, or neither.
This recommendation can be directly adopted or adjusted by the Customer Care team.
51% of customers will never do business with a company again after just one bad service experience.
Source: info.microsoft.com
Results
The key results include:
More efficient processing
By automating the classification of inquiries and introducing specific summaries, processing time was significantly reduced.
Optimized processing order
Sentiment analysis enables optimized prioritization of inquiries, leading to more timely solutions.
Smooth growth
Sunny Cars can continue to grow without the disadvantages of hiring processes (lengthy search, onboarding, and offboarding of highly qualified employees).
Cultural change
- A ‘AI-makes-sense’ culture is emerging – employees are open to the use of AI and are increasingly discovering needs and use cases.
- Questioning the thought ‘We’ve always done it this way’ thanks to visible impact and the feeling of relief.
Summary and key takeaways
This case study shows how artificial intelligence in customer service can greatly benefit both employees and customers. By implementing AI-based solutions, Sunny Cars was able to clearly optimize the customer care process and the customer experience.
Key takeaways for companies looking to implement AI in customer service:
- Yes, AI can strongly support customer care teams in various ways.
- Starting with the ‘low-hanging fruit,’ the easily achievable successes, allows employees easy access to AI.
- Small projects enable immediate experience of the positive impact of AI systems and increase commitment from customers and employees.
This more customer-oriented service handling by Sunny Cars offers the company’s customers even more efficient processing of their concerns. The company plans to further expand and refine these AI-based approaches.