United StatesTechnology and telecom2022 – ongoingAT&T
Real-time machine learning models to detect wireless fraud
AT&T cut fraud attacks by up to 80% with real-time machine learning models on its wireless network
Problem
AT&T serves 182 million wireless customers and faces scams such as robocalls, robotexts, illegal phone unlocks and identity theft. Its fraud data sat in many separate systems, and detection was not fast enough to stop fraudsters in the moment.
Approach
AT&T's data team ran more than 100 fraud detection models in production on Azure Databricks. The models took in structured and unstructured data from many systems and worked in real time. They sent automatic alerts and recommendations to call center, retail and online staff.
Results
Takeaway
AT&T sent model alerts straight to call center, retail and online staff, the people who deal with customers. The company did not say how much fraud cost it saved.
Categories
- Size
- Large enterprise
- Industry
- Technology and telecom
- Business function
- Finance, risk and legalCustomer service
- Innovation type
- Process innovation
- AI technology
- Predictive analytics and ML
- Adoption
- Off-the-shelf tool or platform
- Evidence
- Vendor case study
Sources
Source summaries
Databricks describes how AT&T runs more than 100 fraud detection models on Azure Databricks. The story reports an 80% decrease in fraud attacks (stated as up to 80% in the body text) and millions of dollars saved in potential fraud costs. AT&T did not give an exact dollar figure.
Full details are in the original sources.