Banks used to build their own rules-based fraud-prevention engines, seeing vendor software as an expensive, one-size-fits-all, second-best option. But the increasing sophistication of the fraudsters has changed all that and they are rightly turning to software that incorporates artificial intelligence.
Some are trying to build their own AI-based software. However, a number of reasons including cost, efficacy and speed of both implementation and throughput prove that it is better to partner with a specialist third-party vendor.
Analysis of the total cost of ownership for AI-based fraud-prevention applications built in-house compared with that from Swiss FinTech NetGuardians reveals a startling truth –NetGuardians comes out 77 percent lower cost.
But the advantageous price differential is only part of the story. The reality of today’s fast-moving fraud landscape is that banks need more data than they alone can collect to build and feed their algorithms to keep them up to date. This means building in-house is not only more expensive but is also likely to be significantly less effective.
Fraud prevention demands speed – speed implementing the software so that it can quickly start doing its job, speed when processing transactions and speed at spotting new fraud types so they can be stopped. Without doubt, a pre-integrated solution is going to be up and running quicker than building one in-house and any solution worth its salt will operate in real time and be fully scalable to cope with all the transactions of even the largest Tier 1 bank.
This should be enough to give a bank that has spent years trying to build an effective AI-based fraud-prevention platform pause for thought. Indeed, one NetGuardians’ customer abandoned building a solution after three years of trying. Implementation of our solution took just three months, and the bank is able to build on top of our AI platform. This allows its data science team to use the NetGuardians’ blueprint and best practices rather than having to develop their own from scratch.
Trying to build its own AI solution had not just wasted internal resources, but the project had also left the bank vulnerable to fraud, with its own in-house software capturing less than 30 percent of fraud and adversely affecting the customer experience. A large – and growing – number of false alerts had meant customers were contacted too often for comfort. The in-house system was neither sensitive enough to capture fraud nor clever enough to read customer habits.
With customers putting an ever-bigger emphasis on service, getting it right is vital in today’s competitive market. To this end, NetGuardians uses AI to build up highly accurate customer profiles so transactions that are out of character can be spotted with a very high degree of accuracy. In addition, we pool data, as permitted, from other banks, grouping similar bank customers together to fine-tune our models – something a bank working alone can never do.
As a result, banks using our software typically see a drop in false alerts of 85 percent – that’s a lot fewer unnecessary calls to customers. Another benefit is that our software can be multi-tenanted. This means a global bank operating in the US, Europe and UK can use one application across all three entities and ringfence data so that it can be stored and accessed to comply with different jurisdictional regulations. This makes management and maintenance much easier. Developing multi-tenancy software in-house has time and again proved impossibly complex.
But it’s not just about cost, speed and efficiency, important though they are. Bringing in a specialist can also help with the culture change banks face when moving from a rules-based fraud-prevention approach to a more accurate one centered around AI.
Those working in banks in fraud prevention have typically been trained in a rules-based approach and despite its increasingly apparent shortcomings, can be quite unsettled about switching to AI – whether it’s built in-house or not. A well-structured change management program helps them adjust. At NetGuardians, we have worked with more than 80 banks worldwide, helping them teach their staff to trust AI.
While some of the trust comes from spending time on training, it is also important that staff understand why an alert has been raised. NetGuardians software gives a clear explanation with each alert as to what was suspicious about the transaction – the timing, the amount, the payee, the location, perhaps the browser used. We also recommend that to begin with a bank runs our software in parallel with its existing rules-based engine. Once a degree of trust has been established, the next step is to start to reduce the rules, keeping only those that are key for compliance. Finally, the bank can keep minimal rules and rely mostly on the AI software. Indeed, drawing on our explainable AI, we have developed a blueprint for this important trust building exercise.
In today’s market, where fraud types are constantly evolving and where scale and experience count more than ever, banks are leaving themselves exposed if they try to build an AI-based fraud solution from scratch. Far better to build on top of a proven package from a specialist as it would do when buying Oracle, Microsoft or SAP for Enterprise Resource Planning (ERP) needs. The result will be less fraud and lower costs. You can bank on it.
Michaël Gingins is Head of Operations & Professional Services.