Some $442bn has been lost globally to financial fraud in the last year and as banks around the world accelerate their shift to data-driven fintech models, the complexity of these systems is creating new blind spots for fraud.
As Michael Down tells RBI, fraud now moves through coordinated networks of mule accounts, synthetic identities, and transactions, rather than in isolation.
Go deeper with GlobalData
Access deeper industry intelligence
Experience unmatched clarity with a single platform that combines unique data, AI, and human expertise.
As AI is helping to increase the speed, scale and sophistication of attacks, the real-world impact is one in which financial firms aren’t currently set up to connect the dots within these invisible networks. And he explains how graph intelligence goes beyond detection, to change what’s operationally possible for the teams behind the scenes and is successful in uncovering suspicious patterns that traditional systems miss.
RBI: In your opinion, how is AI increasing the speed, scale and sophistication of attacks in the FS sector?
Michael Down, Global Head of Financial Services at Neo4j:
AI has fundamentally changed the economics of fraud. What once required coordinated human effort – creating fake identities, crafting convincing phishing messages, and manipulating documents – can now be automated and deployed at scale in minutes with AI.
We’re seeing this play out across multiple fronts: deepfakes are used to bypass identity verification systems through cloned voices, manipulated video and facial imagery. Separately, AI-generated synthetic identities combine fabricated personal information and manufactured digital footprints to create accounts that appear entirely legitimate, while AI-powered phishing campaigns continuously test and refine the approaches most likely to succeed. The result is fraud that is more convincing, more coordinated, and more sophisticated than anything the industry has seen before.
The challenge for financial institutions is that legacy detection systems were built for a different threat model – one where anomalies appeared individually and could be caught through rule-based controls. Today’s AI-enabled fraud deliberately spreads activity across accounts, devices, and institutions to avoid triggering those rules. Traditional systems, therefore, need to adapt.
RBI: You have referenced work your firm did with BNP Paribas Personal Finance to reduce fraud by 20%…can you give some information on how the solution achieved this?
Down: We helped BNP Paribas Personal Finance fundamentally shift from treating fraud as an isolated event to helping them identify connected fraud networks.
The bank processes a staggering 800,000 credit applications a year, and while its existing systems could run rule-based checks and maintain blacklists of known fraudsters, they couldn’t see how data insights were connected. And that’s exactly where sophisticated fraud thrives. From shared devices across unrelated accounts to coordinated application timing, the worrying patterns were there, but they were effectively invisible because the data sat in silos.
By implementing Neo4j’s graph technology, they created a single connected view of customers, accounts, and devices. That meant hidden networks became visible as soon as a new application came in. When a fraud signal appeared, investigators could immediately trace related entities and uncover the wider network, rather than piecing it together manually.
In practice, the bank achieved a 20% reduction in fraud, alongside significant time savings for investigators. Analysts could see the full picture in one place, enabling them to catch fraud earlier and reduce the effort required to build a case.
For me, that’s where the true results lie. Graph intelligence goes beyond detection, to change what’s operationally possible for the teams behind the scenes.
RBI: I read that Neo4j is already working with thousands of organisations — from government agencies to Fortune 500 companies — How significant is the FS sector to the firm as an industry vertical?
Down: Financial services is a significant vertical market for Neo4j, as AI continues to completely reshape the fraud landscape. In banking, insurance, and capital markets, every transaction, customer interaction, and regulatory obligation exists within a web of connections. That makes financial services a natural fit for graph technology.
A company might start applying a connected data model to fraud detection, then extend the capabilities to regulatory impact assessment or customer analytics. Graph technology isn’t a single-purpose tool – once deployed, organisations find new problems to apply it to as the connected view of data unlocks new use cases. As financial crime becomes more networked and compliance demands greater transparency, the value of that connected view only increases.
Neo4j works with major global organisations including Citigroup, UBS, and Zurich Insurance Group, on various use cases, typically fraud detection, anti-money laundering, regulatory compliance, and risk management. We also work with other innovative companies such as Daimler Truck and Uber on their AI initiatives whereby the accuracy of the relationships in their data directly determines how well AI models can detect patterns, flag anomalies, and generate reliable outputs.
RBI: The Neo4j solution, is I assume, compatible with all of the main cloud solutions that banks deploy eg AWS, Google, MS?
Down: Yes. Neo4j runs natively across AWS, Google Cloud, and Microsoft Azure through Neo4j Aura, our fully managed cloud database service. Institutions can deploy in whichever environment that fits their existing infrastructure, security requirements, and data residency obligations.
This matters because financial services organisations rarely operate in a single cloud environment. They have legacy systems, multi-cloud strategies, and strict governance around where data resides and how it moves. Neo4j is designed to work within that reality rather than against it.
Beyond cloud compatibility, we integrate within the broader data ecosystem that banks already rely on, including business intelligence tools, machine learning pipelines, and streaming platforms. The goal is to add a relationship intelligence layer to existing infrastructure, not replace it. Whether an institution is running workloads on-premises, in a hybrid model, or fully in the cloud, Neo4j fits into that architecture and delivers graph capabilities wherever the data lives.
RBI: Looking ahead, what are Neo4j’s key priorities, growth targets, or strategic ambitions within the financial services sector over the next 12–24 months?
Our priorities in financial services over the next 12 to 24 months are clear: go deeper on fraud, expand into risk and regulation and position graph technology at the heart of the generative AI transformation underway in financial services.
On fraud, we want to move organisations beyond point solutions. The opportunity for meaningful change is a holistic Financial Crime Graph Intelligence Platform that connects the full fraud lifecycle – from detection and investigation, through to regulatory reporting and feeding back insights for model retraining. That closed loop is where enterprise-grade fraud detection becomes truly powerful.
Risk and regulation is a natural extension of that. We are seeing strong growth in our partner ecosystem here, and we are sharpening our own capabilities to better support customers with connected approaches to compliance and resilience.
As graph technology becomes more foundational to Generative AI, we’re exploring more opportunities to help banks build digital twins of their enterprises, support agentic memory through context graphs, and power use cases like fraud detection and Customer 360 where relationships in data matter most.
