Q1. Could you start by giving us a brief overview of your professional background, particularly focusing on your expertise in the industry?

I work at the intersection of fraud, payments, and AI for banks, PSPs, and regulators in MENA. My focus is on Agentic AI and Fraud Management as a Service, where I help institutions move from legacy rule engines into cloud and AI-powered fraud platforms that are acceptable to regulators and practical for operations teams.

Over the last years, I have worked with issuers, acquirers, and payment processors across the region on card fraud, e-commerce risk, and digital wallets. A big part of my work now is designing FMaaS and RMaaS operating models, AI agents for fraud and FP&A, and ISO 42001-aligned AI governance so that CFOs, CROs, and regulators can trust the models, not only the KPIs.

Alongside product and strategy work, I spend a lot of time training finance and risk teams on GenAI, agentic AI, and data-driven decision-making, so they can own the roadmap rather than rely fully on vendors.

 

Q2. What typical ROI multiples have banks seen from FMaaS vs. in-house fraud systems, and what false positive reductions drove those results?

In my experience, when FMaaS programs are set up and run well, they usually pay back at least three times what goes in over about three years, which is a lot more than you typically see from just improving in-house tools. The big wins come from actually cutting fraud losses and significantly reducing false positives, so teams waste less time and customers have a smoother experience. There isn’t a single global benchmark everyone cites, but this pattern has been quite consistent in the implementations I’ve been involved with.

Public studies on AI fraud platforms, many delivered as SaaS or managed services, show:

• A meta-analysis of 47 banking studies found that AI fraud systems detect 87–94% of fraud cases while reducing false positives by 40–60% compared with traditional rules-based frameworks. 

• Large bank case studies show up to 60% fewer false positives, plus much faster and more accurate fraud detection. 

• Some vendors report about a 65% reduction in fraud losses within the first year, around 40% fewer false positives, and roughly 3x ROI thanks to both loss avoidance and operational savings. 

FMaaS builds on the same economics, then adds:

• • Lower up-front costs and faster rollout, so banks start seeing benefits sooner

• Shared threat intelligence and model updates across multiple institutions, which improves hit rates and supports maintaining ROI over time

In my own projects, here are the practical benchmarks I rely on:

•  Around 3x ROI over three years for a bank moving from a basic rule engine—this is a conservative estimate

• About 5x as strong performance, where there was a lot of fraud and manual workload to fix

These ranges come from real implementations, not a single industry benchmark, so they shouldn’t be read as a one-size-fits-all standard. That said, the practical takeaway is straightforward: for FMaaS to be financially worthwhile, banks usually need to see at least a 40–60% reduction in false positives, while maintaining or even improving their fraud detection performance.

 

Q3. How does cloud FMaaS typically cut bank capex vs. on-premise solutions, enabling regional scaling in high-growth markets?

Cloud FMaaS works well in practice because it turns what would normally be a big upfront capital spend into a more predictable ongoing operating cost, while also taking advantage of the broader cost efficiencies that cloud has brought to financial services.

Research on cloud banking and data platforms shows:

• Moving data and analytics workloads from on-prem to cloud can reduce run-rate costs by about 40 percent and eliminate large hardware upgrades, mainly through lower support and infrastructure costs. 

• Studies on cloud in financial services report 20–30 percent lower total IT spending and up to 25 percent lower infrastructure costs for institutions that adopt cloud-native architectures, with much faster time-to-market. 

FMaaS uses these economics in a focused way:

• The bank avoids buying and maintaining fraud hardware, databases, and commercial software

• The provider runs a multi-tenant fraud stack, spreads infrastructure and model costs across many clients, and charges per transaction, per account, or per bundle

• With elastic capacity, a single FMaaS stack can support new corridors, wallets, or geographies without requiring each bank to undergo another capital expenditure cycle.

In fast-growing markets, being able to scale up quickly often matters more than squeezing out every last bit of cost savings. With FMaaS, regional or pan-African banks can roll out fraud protection for new countries or partner PSPs in just weeks—instead of the months or longer it takes for on-premise setups—while still keeping the governance controls needed for outsourcing and cloud.

 

Q4. Which partnerships and regulatory changes boosted FMaaS margins (15–25%) and market penetration in emerging regions?

I haven't come across solid public data that pins FMaaS providers' margins consistently at 15–25%. That said, I've seen partnership models and regulatory changes that make those margins very achievable for efficient operators.

On the partnership front, three themes really stand out:

• Network or switch-level fraud utilities: In places like Indonesia, payment switches and processors team up with fraud vendors to deliver shared fraud management for banks and fintechs. This creates a multi-tenant platform with lots of users, which drives up utilization and spreads out the fixed costs. 

• Cloud and AI alliances: More and more banks are combining local FMaaS specialists with global cloud and AI platforms. These alliances help FMaaS vendors access cheaper infrastructure, use built-in AI services, and integrate more easily with modern core systems—leading to better unit economics. 

• Ecosystem VAS partnerships: Wallets and PSPs need add-on services like chargeback analytics, merchant risk scoring, and credit decisioning. FMaaS providers who bundle these extras capture more revenue per transaction.

On the regulatory side, the big accelerators in emerging regions are:

• Clear outsourcing and cloud frameworks for critical services, such as SAMA’s guidance on cloud and outsourcing in Saudi Arabia and similar work in Europe and other regions. These frameworks define how banks can place fraud monitoring in the cloud while remaining compliant 

• Stronger AML and financial crime expectations that encourage advanced analytics and utilities, which make shared FMaaS and RegTech providers more attractive to both banks and supervisor

When these factors come together, FMaaS providers can achieve strong margins by operating a single platform for multiple clients, reusing models and processes across markets, and adding value-added services on top of core fraud monitoring. The actual margin percentage differs widely by firm, so I do not consider any single figure to be standard.

 

Q5. What GenAI advantages outperform traditional ML in FMaaS, especially for cross-border transactions?

GenAI does not replace traditional machine learning scoring in FMaaS. Instead, it complements these models and addresses gaps, especially in managing unstructured data, improving explainability, and managing multi-jurisdictional requirements.

For cross-border transactions, GenAI and agentic AI add three important advantages:

Richer investigations and narratives

• GenAI can read KYC files, transaction notes, chat logs, and external intelligence, then produce a structured case summary or SAR narrative in minutes instead of hours for an analyst. 

• This reduces handling time and helps banks keep up with the volume that real-time cross-border rails create

Better multi-corridor context

• Cross-border payments involve multiple jurisdictions, sanctions lists, and payment schemes. Agentic AI can orchestrate calls to sanctions data, network graphs, device intelligence, and corridor risk models, then combine them into one explanation for a decision. 

• Traditional ML models score transactions, but they do not naturally generate narratives that regulators and customers expect.

Language and interface advantages

• GenAI models handle multiple languages and scripts, which is critical when remittance flows and trade corridors involve Arabic, English, Asian, and African languages in the same network

• They also power chat-like analyst tools that let risk teams ask questions in natural language across millions of transactions

At the core, risk scoring continues to depend on supervised and unsupervised machine learning. The GenAI layer makes these scores more usable, auditable, and scalable in day-to-day operations, which is critical for FMaaS success.

 

Q6. What level of wallet user adoption (for example, over 25%) has been linked to doubling the growth of value-added services, compared to lower adoption rates (under 15%) where growth stalled?

I have not seen rigorous public research that proves a specific link, such as “over 25 percent wallet adoption doubles VAS growth compared with under 15 percent”. I cannot confirm that exact relationship from published data.

What we do know from market studies is that:

• Digital wallet users worldwide are expected to exceed 6 billion by 2030, more than three-quarters of the global population, and in this environment, value-added features such as BNPL, virtual cards, and digital identity are described as vital for differentiation.

• Payment research on value-added services notes that wallets and card programs use data-driven rewards, instalments, and similar services to create new revenue streams once basic payment functionality becomes commoditized. 

In practice, I use simple adoption bands as a management tool, not as a scientific rule:

• Below roughly 10–15 percent of active customers are using the wallet, VAS is mostly in pilot mode, and growth depends on marketing spend.

• Somewhere around the 20–30 percent active user band, daily usage is high enough that instalments, offers, and small VAS lines start to matter in P&L.

• Beyond that range, superapp-style economics can start to appear, once engagement and data density are strong.

These bands are planning heuristics for product and FP&A teams, not statistically proven thresholds. Any serious investor or bank should validate them on their own data before linking them to bonus targets.

 

Q7. If you were an investor looking at companies within the space, what critical question would you pose to their senior management?

My main question would be:

Show me, corridor by corridor and segment by segment, how your FMaaS platform changed three things over at least twelve months for live clients: net fraud loss rate, false positive rate, and regulatory findings, compared with the previous control stack, and explain what protects those gains when fraud patterns shift and when you enter a new market.

Everything else flows from that.

If they can answer with audited numbers, clear before-and-after baselines, and a repeatable playbook for new markets, then the business has real substance. If the answer is mostly roadmap and marketing language, the FMaaS story is not yet investment grade.