ChinaFinance and insurance2018 – ongoingPing An Property & Casualty Insurance
Credit-based smart auto claims
Ping An let drivers settle minor car claims from photos, with AI damage assessment it put at 95% accuracy
Problem
A conventional auto claim at Ping An P&C went through an average of 5.18 manual steps per case, including investigation, loss assessment, review, document collection and payment. Minor accidents took as much handling as larger ones.
Approach
As part of a transformation launched in 2018, Ping An P&C applied image recognition to assess car damage from customers' photos, together with text recognition for repair bills and face recognition for identity checks. In January 2019, after a trial in Guangdong, it rolled out nationwide a credit-based scheme: drivers receive a claims quota based on driving behaviour and credit history, report an accident in the app, upload photos and confirm the payout online, while fraud cancels the quota.
Results
Takeaway
The scheme removed steps by trusting low-risk drivers within a quota, with image recognition doing the assessment and the quota limiting the loss when it was wrong. The accuracy and speed figures are the insurer's own and were not independently tested.
Categories
- Size
- Large enterprise
- Industry
- Finance and insurance
- Business function
- Customer serviceFinance, risk and legal
- Innovation type
- Process innovationProduct and service innovation
- AI technology
- Computer visionPredictive analytics and ML
- Adoption
- Built in-house
- Evidence
- Company's own claim
Sources
Source summaries
Ping An's release announces the nationwide launch of a credit-based auto claim scheme after a test in Guangdong. Drivers report and settle claims on a phone against a quota tied to driving behaviour and credit. It states that 30% of cases already used AI image recognition for loss assessment at 95% accuracy, against an average of 5.18 manual workflows per case in the traditional process.
The trade outlet reports the launch of Ping An P&C's credit-based claim scheme, in which compensation comes from a dynamic quota linked to driving history. It repeats the company's accuracy figure of up to 95% and quotes deputy managing director Zhu Yougang on the aim of encouraging safer driving.
Insurance Business reports Ping An executives' account of AI in motor claims, where image recognition prices damage within seconds and a trust-based settlement takes four steps and around three minutes. The insurer removed investigation, loss assessment, review and several other steps from its own procedures as part of a transformation begun in 2018.
Full details are in the original sources.
