Data-sharing APIs make anti-fraud measures a commercial opportunity for CSPs

Commercial context

Despite the vital importance of effective fraud detection to organizations and customers across most industries, the uncomfortable reality is that fraud is on the rise. Fraudsters are increasingly sophisticated, and evolving rapidly with the growth of 5G and new connected services. The reality is, fraud is now far more complex than it ever before – it is also far more organized and underpinned by evolving business models to match. Conventional defences are being tested to the limit, and in some key respects, are struggling to keep pace. The consequences are profound: the estimated total loss to fraud in the telecoms sector was US$38.95 billion in 2023 – a 12% increase on 2021 – representing 2.5% of telco revenues worldwide. A major proportion of this increase has come from device fraud, which at 27% of total telecoms fraud losses are now exceeding US$10.8 billion per year - with 29% of device users having experienced account takeover in 2023.

Yet even these deeply troubling figures are only a fraction of the total losses to fraud last year in the financial services sector, which saw an annual loss of US$485.6 billion. Clearly, traditional fraud investigation methods are insufficient, with both CSPs and financial institutions now at heightened risk. A particular barrier to effectiveness is that current industry solutions lack standardization, limiting scalability and interoperability; financial institutions also lack ready access to live data necessary for detecting suspicious transactions quickly.

The solution

The AI-powered fraud defense: Turning vulnerability into profitability Catalyst is developing an AI-based virtual agent which uses machine learning, natural language processing, generative AI and data analytics to identify and comprehend patterns of fraud to improve detection processes. By automating and streamlining labor-intensive processes, such as case investigations and creation of comprehensive reports on risk or patterns of suspicious activity, this intelligent agent helps reduce resolution times by 20-40%, fraud losses by 15-30%, and operational costs by 15-25%. Importantly, it also allows reallocation of human analyst efforts to higher-value tasks.

The Catalyst also proposes a new fraud management API that will extend current capabilities by enabling CSPs’ partners to tap seamlessly into their datasets to improve generation of risk scores while enhancing reusability and integration with other services. The framework will adhere to TM Forum standards and Open Digital Architecture, which are crucial to ensuring secure and compliant sharing of data, and to fostering trust among stakeholders. The new TMF 770 Fraud Management API will standardize data sharing methods between CSPs and financial institutions and other key verticals, with the pilot use case being the provision of credit score-related data to financial service providers.

The API will enable solutions like the pilot AI agent to draw on indicators from CSP data sources such as account information, device identity and SIM tenure to identify potentially risky transactions, such as those soon after a SIM change, or where there appears to be call forwarding in place (which may indicate hijacking of the line by fraudsters). Network authentication, live call verification and geofencing will also serve as key risk management capabilities.

Application and wider value

This Catalyst shows a clear route to an industry standard for AI agent-based fraud prevention which can genuinely transform how we combat fraud. Agents’ ability to continuously learn and predict suspicious activity in real-time provide a much-needed means to halt fraud before it escalates. Underpinned by unique insights which only CSPs can provide, agents ultimately make it possible to provide financial institutions with crucial real-time data, enhancing their risk assessment capabilities and overall fraud prevention strategies

AI agents enable CSPs to fight fraud outside of the telecoms industry and effectively support any sector in need of a modern defense. In doing so, CSPs can ethically monetize valuable data insights, unlocking new revenue streams and strengthening commercial relationships with the financial services sector. In practice, as the project has shown, this can take many shapes and forms. For example, AI agents can enable automated internal fraud investigations, helping to expedite identification of key scenarios such as handset fraud, and dramatically reduce the time needed to resolve them. They can also make computation of risk scoring far quicker for financial institutions, and more accurate by drawing on valuable data which is currently either difficult to access or entirely unavailable. The more complete view of users made possible through this data is expected to support a reduction in loan defaults by 10-15%.

Perhaps of most lasting value, however, is the potential to support mining of patterns and detecting of anomalies in CSPs’ datasets via advanced analytics. Through deep scrutiny of abnormalities, CSPs and their partners can benefit from generation of actionable insights that can inform long-term prevention strategies, switching the focus from identification and remedy of actual fraud to avoidance by closing off avenues to fraud as they become clearer. The project team anticipates that proactive detection and prevention in this way should reduce fraud losses by 30-50%, which in turn fosters trust, customer satisfaction and improved customer retention.

The solution provides considerable, mutual benefits to participating CSPs and financial service providers far beyond the immediate wins in combatting fraud to reduce revenues losses and protect customers. The project will more broadly foster secure, collaborative data-sharing practices, breaking down silos and providing deep learning experiences which will benefit both industries as the era of sophisticated digital services develops. It will also enable efficiency savings from reduced operational costs by streamlining traditionally manual processes, and support improved decision-making and scalability more broadly across participating organizations as they grow increasingly adept at drawing on the APIs to deploy automated solutions of this kind.

“This project benefits telcos by reimagining fraud management, allowing them to focus on complex initiatives while AI agents handle apparent fraud cases and share information with stakeholders efficiently and promptly,” explains Carlos Celaya, International Fraud and Revenue Assurance Manager at Orange. “Additionally, it creates new revenue opportunities through APIs, benefiting not only telcos but also the BFSI sector. The ultimate goal is to foster a more secure world with less fraud, where digitalization, data, and AI are harnessed for the greater good of society.”