FOCAL x Google Cloud: Speed & Compliance in AI Fraud

As financial institutions navigate a rapidly evolving threat landscape, AI and cloud technologies have emerged as crucial tools in the fight against fraud. In April 2025, FOCAL and Google Cloud co-hosted a thought-provoking webinar titled “AI-Driven Fraud Detection in Financial Services: Balancing Speed and Compliance.” Featuring Gerrit Bekker, Director of Data Science at Mozn, and Abdulrahman Alshaker, Digital Native Lead at Google Cloud, the session offered deep insights into how modern infrastructure, machine learning, and real-time analytics are transforming fraud prevention.

Watch The Full Webinar

The Role of Cloud Infrastructure in AI-Driven Fraud Detection

A foundational theme of the webinar was the strategic role of cloud infrastructure in enabling AI-driven solutions. Abdulrahman emphasized that cloud-native environments allow faster development, scaling, and deployment of fraud detection models, particularly when handling massive volumes of real-time financial data.

“Cloud infrastructure is not just about storage. It enables integrated data processing, accelerates AI pipelines, and ensures compliance through built-in security and encryption,” — Abdulrahman, Google Cloud

Key takeaway: Cloud services such as Google Cloud’s Vertex AI and BigQuery enable institutions to build and iterate fraud detection models with greater flexibility and security—reducing time to market and improving real-time responsiveness.

What Is Modern Fraud? Evolving Threats in Financial Services

Gerrit Bekker, a veteran in risk systems across GCC banks, provided a compelling overview of what modern fraud looks like. Today’s threat actors are highly organized, often using AI themselves to launch sophisticated, adaptive attacks.

Common challenges discussed:

• Synthetic identity fraud

• Money mule networks
  

• Insider threats and social engineering
  

• Real-time digital transaction fraud
  

“We’re no longer just reacting to fraud. We’re predicting and preventing it using AI,” — Gerrit Bekker, Director of Data Science at Mozn

Insight: Traditional rule-based systems are not sufficient. Financial institutions need adaptive, learning-based models that evolve with attacker behavior.

Read more: Transaction Fraud Detection: Top 8 Tips and Best Practices

Building AI Models for Fraud Detection: Data, Governance, and Compliance

Developing reliable fraud detection models requires clean, contextualized data—something the cloud helps streamline. The speakers discussed best practices around data governance, labeling, and cross-functional collaboration between engineering, compliance, and risk teams.

Gerrit also highlighted the importance of:

• Data anonymization
  

• Synthetic data generation for model training
  

• Model explainability for audit and compliance reporting
  

“Explainable AI is no longer optional in regulated industries. Regulators want to see how the model makes decisions,” — Gerrit Bekker

Real-World Examples: AI in Action

Several use cases were mentioned during the webinar, showcasing how AI-driven systems have been deployed in production:

• GCC Banks using real-time ML pipelines to flag suspicious transactions before approval

• Cross-industry fraud intelligence sharing platforms, enhancing predictive detection via anonymized datasets

Notable stat: FOCAL has proven to be a successful tool for one of its key clients. After deploying our solution, we observed a significant improvement in fraud detection capabilities. Specifically, the transition from traditional human-defined rules to machine learning-based models resulted in an 80% reduction in fraudulent activity. This marks a substantial enhancement in both efficiency and accuracy, highlighting the value of data-driven automation in real-world fraud prevention.

Future Trends: Where AI and Cloud Are Headed

When asked about future developments, the panelists discussed self-learning fraud detection systems, multi-cloud data ecosystems, and privacy-preserving AI techniques such as federated learning.

“We anticipate more institutions will adopt privacy-first AI approaches to comply with global data laws while still benefiting from collaborative intelligence,” — Abdulrahman Alshaker

Trends to watch

• AI Ops and automation for fraud response
  

• Cloud-native compliance tooling (e.g., policy tagging, data lineage tracing)
  

• Increased use of generative AI for customer verification and case triage

Watch The Full Webinar

Q&A Highlights: Speed, Compliance, and Practical Deployment

During the Q&A, both speakers answered practical questions from the audience:

Q1: How do institutions balance speed and compliance?

Abdulrahman noted that institutions don’t have to sacrifice one for the other. With cloud-native architectures, data is processed at speed while maintaining full encryption and audit trails, helping meet both operational and regulatory demands.

Q2: What KPIs should be used to measure AI-driven fraud detection systems?

Gerrit outlined several:

• Number of early fraud interventions
  

• Volume of real-time detection alerts

• Reduction in manual investigation workload
  

• Percentage of fraud cases intercepted before money movement
  

Q3: What are the main challenges in deploying AI in high-risk markets?

Both speakers acknowledged:

• Limited access to labeled data
  

• Regulatory complexity

• Infrastructure readiness Gerrit also emphasized that false positives and poor-quality data are persistent hurdles.
  

Q4: How does Google Cloud compare to others in this space?

Abdulrahman explained that Google Cloud focuses heavily on secure-by-default infrastructure, and that its AI tooling is integrated natively with its security and compliance architecture. This helps financial clients scale while staying within local and international regulations.

Q5: Can smaller banks or fintechs start using AI without large budgets?

Yes. Abdulrahman said Google Cloud offers tools that let teams build and deploy models without large engineering teams. He also pointed out that startups can access pre-trained models and use usage-based pricing to avoid high upfront costs.

Q6: How is data encrypted and protected?

All data on Google Cloud is encrypted at rest and in transit by default. For sensitive data, additional layers of encryption and access controls can be configured to meet strict internal or jurisdictional requirements.

Final Thoughts

This webinar reinforced a critical message: fraud is getting faster, smarter, and harder to detect, but so is the technology fighting it. Financial institutions must rethink how they approach fraud detection, moving from siloed tools to integrated, AI-powered platforms that work in real time.

The combination of cloud-native infrastructure, predictive machine learning models, and regulatory-friendly architecture, offers a new playbook for institutions looking to balance speed and security in the digital age.

Watch Webinar Replay