GetGo builds an AI model on OCI to help it review images for vehicle damage
The carsharing service uses OCI AI Infrastructure to assess up to 50,000 daily photos for vehicle damage, working 100X faster than human reviewers.
“What we gained by developing our AI model on Oracle Cloud Infrastructure was the speed at which we could process and act on huge amounts of data and images that we collect daily. Without the model, it would take us weeks or months to process the same amount of data we can review in a day with AI.”
Owning a car is very expensive in Singapore, so only a third of households have their own vehicle. As Singapore’s leading and largest carsharing and short-term car rental service, GetGo Technologies allows users and their families to access private mobility through its growing fleet of 3,000 cars across 1,700 locations on the island. Daily, the company receives up to 50,000 user-submitted photos of its vehicles, making manual review of images to check for damages impractical. To address this, GetGo developed an artificial intelligence image processing model on Oracle Cloud Infrastructure (OCI) that assesses all those images in a few hours, completing what would take staff weeks to review. Trained on OCI to detect damages, the AI model helps GetGo respond quickly and accurately to damages, provide timely feedback to drivers, and get cars repaired and back in the fleet faster, which improves service quality and customer retention.
The Oracle AI Customer Excellence Center really helped identify the match between problems and solutions, so we’re not only focusing on what we could buy but what we could solve.
Why GetGo chose Oracle
With up to 7,000 users each day, GetGo’s carsharing service lets customers register, book a car, and get driving in minutes. Its business model is fully digital and self-service, letting drivers hit the road without human assistance.
As part of the rental process, drivers submit at least eight photos of the vehicle: four before they drive and four when they return it. Before using AI, GetGo would review photos only when a customer reported a damaged vehicle, triggering an employee to find the time and date of recent rentals to match up and compare photos. But that process was too slow and reactive, and the company struggled to provide timely feedback to users and accurately identify when and how damages occurred. Reviewing all photos manually wasn’t an option: existing staff would need 25 workdays to assess just one day’s images. The company previously explored third-party image processing services, but learned that they would cost 50 cents per image.
Instead, the company chose to collaborate with Oracle, using OCI AI Infrastructure to develop and run an AI-powered image processing model that could rapidly and cost-effectively detect vehicle damages. GetGo and Oracle co-created solutions specific to GetGo’s challenges. The company’s in-house data science team worked closely with the Oracle AI Customer Excellence Center, which provided technical guidance, training, and resources to help GetGo develop the AI model, get it into production, and run ongoing inferencing on OCI.
OCI helped GetGo move its AI model development from problem identification to production in less than three months.
Results
GetGo gained significant improvements in speed, accuracy, and cost after adopting OCI AI Infrastructure. The company’s AI model is trained and deployed on OCI and integrated with the company’s larger infrastructure. This allowed GetGo to process roughly 50,000 daily customer images up to 100 times faster than manual review. It continues to train the model and annotate damages so that detection accuracy constantly improves, which helps the company avoid wrongful damage claims against customers.
The AI model assisted GetGo’s shift from a reactive to a proactive approach in managing its vehicle fleet, helping it improve efficiency, user experience, and service quality. The company can quickly make vehicle repairs and provide accurate and timely feedback to users. OCI also let GetGo manage infrastructure expenses effectively and allowed for scalable processing with tight cost control.
By automating damage detection in user images, GetGo’s team members were freed to focus on higher-value work. The company views AI and its technology as a tool to elevate employee capabilities, moving them from routine tasks to more strategic, creative, and customer-focused roles, including designing the logic behind how photos get flagged.
OCI provided a flexible platform for GetGo to explore additional AI applications, with the damage detection model serving as the first step in the company’s AI innovation efforts. GetGo sees AI as a key tool to help solve complex operational challenges and improve its mobility service. In addition to refining the existing damage detection model, the company hopes to develop AI applications supporting pricing models for repairs and even cleanliness assessments.
About the customer
GetGo is Singapore’s largest carsharing service. It offers users the convenience of private mobility through renting vehicles on its mobile application on a pay-per-use basis without requiring deposits or membership fees.