Tips for researching an AI-based, AML/fraud detection proof of concept
BY Deleep Nair
Not all anti-money laundering and fraud solutions are created equal, and this has caused definite frustration among financial institutions. In fact, according to a 2019 study by Duffs & Phelps, 30% of institutions surveyed rate at least one of their AML components as being either “not at all” or only “somewhat” effective.
Even though AI is gaining traction in financial crime detection, there are vast differences among platforms. Many well-capitalized start-ups and consulting organizations simply apply open source to massively complex problems, without the rigor and discipline that enterprise scale and explainability demands.
Adding to that frustration, compliance costs are rapidly increasing. According to a 2022 study by LexisNexis Risk Solutions, the worldwide cost of financial crime compliance is estimated at $274.1 billion, up from $213.9 billion in 2020. The same study finds that the average annual compliance cost per organization has risen by double digits since the start of the pandemic. This is hardly surprising, given the inability of old transaction monitoring systems to find crime without more than 95% false positives, leaving investigators to waste so much valuable time and money. Basically, we have institutionalized inefficiency into our compliance organizations.
Given deficiencies in existing detection systems and the rising cost of compliance, it’s not surprising that an increasing number of banks are reevaluating their existing AML technology. But what is the best and most economical approach to evaluating the best technology and ensuring a comprehensive understanding of a highly regulated market that hasn’t seen true innovation in close to 20 years? We need to be conservative…to compare apples to apples and understand an appropriate move from tech debt incumbered to an effective compliance and FinCrimes function.
One solution: a proof of concept (PoC)
A PoC is by far one of the most effective ways to determine if a particular financial crime detection system is the right fit for your organization. It’s a fail-proof demonstration of the product’s feasibility. You’ll know rather quickly whether or not the system lives up to its promises, saving you time and money. Other benefits include:
- It allows you to work with the vendor during implementation and beyond to customize the system to meet your specific requirements.
- An effective PoC delivers virtually immediate results, greatly reducing false positives to save you time and money, even during the testing phase.
- It provides early indication of future results and compliance costs.
- It provides input into a business case that can justify the project to remove tech debt and de-risk regulatory realignment and burdensome operational inefficiencies.
Some tips for a successful PoC
Tip 1: Ensure that your strategy aligns with your corporate vision
- Be certain that your business case is in keeping with the fiscal, risk mitigation and operational objectives of your bank. This helps quicken the authorization and budgeting process. Align your strategy with selected goals and objectives as described below.
Tip 2: Clearly define the goals and objectives of the PoC
- Have a clear understanding of what you want to achieve with the PoC. This helps guide the selection of the appropriate technologies. Identify specific, measurable business problems that you want to solve, as well as KPIs (key performance indicators) that you will use to measure success of the PoC.
Tip 3: Select the right team
It is critical that you assemble a team of experts with the right skills and experience to conduct the PoC. Include data scientists, software developers, business analysts and subject matter experts from relevant business areas. Also select a project manager to keep the process on track and present results at each stage of the PoC.
Tip 4: Define a clear plan and timeline
Include milestones and deliverables in your PoC timeline. This helps ensure that the PoC is completed on time and that the results are meaningful and actionable. It is important to align the timeline and deliverables with any decisions that need to be made by the organization – budget, technology, personnel decisions, etc. Ask the vendor to assist you in setting significant milestones and deliverables.
Tip 5: Communicate regularly with stakeholders
This ensures that the PoC stays aligned with the needs and goals of the organization. Keep stakeholders informed of the progress of the PoC, as well as of any challenges or issues that may arise. This helps build support for the POC and ensure its success.
Tip 6: Try to use the cloud
It’s cheaper, quicker and focuses on results instead of set-up. Value can be created in weeks rather than in months. While data security will always be front and center, a great detection system doesn’t need sensitive information.
Tip 7: Make sure it’s not an abstract, standalone project
Doing a PoC without the ability to move proven value into BAU would be a waste of time. It needs to be part of a general deployment model. The data used, the business case created, and the next steps should all focus on quickly bringing proven value into the business seamlessly.
Tip 8: Take it seriously
A PoC requires a commitment by both your potential partner and your own team. If handled as the serious proposition that it is, you are likely to be pleased with the results.
The ultimate verification
A PoC is a data point of proof that the product will function as claimed. It identifies any tweaks that may be needed for your bank’s specific needs. It provides decision makers with valuable data to determine the platform’s feasibility and cost effectiveness. It allows you to interact with the vendor in a real-time environment prior to making a longer-term commitment. It is definitely a viable option in your search for a truly effective AI-based AML/fraud detection system.