Decoding the Financial Crime Genome, AyasdiAI, and an Opportunity for the Market
Many have said that little progress has been made in the fight against financial crime. In fact, there is a solid argument that we are regressing:
- Spending more money with fewer results
- Focusing on the symptom of “false positives” rather than the disease itself
- Focused on the rule rather than the spirit of the law, and so undermining the mission
- Excessive focus on regulatory opinion rather than executing on the critical position that banks have in reducing the threats that face us all
The problem with the term “Financial Crime”
Let’s start with why we are doing this. What is the mission, and why should we be passionately focused on solving financial crime?
First, let’s simply define two types of financial crime:
- Fraud and cyber-crime are direct theft from an institution or its customers. It is a clearly defined financial crime with pain that can be directly measured.
- Money laundering, on the other hand, is not theft. It’s the opposite – large cash inflows on profits from or payment for drug trafficking and the narco-economy, pedophile and sex trafficking, rogue sovereign state or terrorist financing. So it’s best to invert the definition to “crime financing” – the facilitation of crimes through enabling financial relationships and transactions.
By overlooking this second definition, we lose what we are trying to stop – human and sex traffickers are exploiting the world’s vulnerable citizens, pedophiles are organized into global communities of exploitation, and terrorists, drug gangs, murderers, and the very worst of humanity.
Prevention of the above falls to the BSA and anti-money laundering teams in the world’s banks. To us, the role of these teams is as impactful to our adversaries as the use of the military or direct kinetic law enforcement.
It is the only function in a financial institution where it is more important to pursue the spirit of the law than just mapping to the rule of law. The problem is large numbers of firms are not doing that, dismissing “crime financing” as merely a regulatory challenge. So, despite huge amounts of investment and organizational equity focused on regulatory reporting, our global banking system continues to be an enormous enabler of heinous crimes.
The tone from the top
Banks, from the board to shareholders, need to show that they are serious about the types of customers they want to service, the pervasive nature of anti-money laundering and trafficking endeavors in their firms, and the gravity of the AML business. This demonstrates a clear acknowledgment of the organizational seniority that focuses on the true crime rather than simply considering success a pat on the head from the regulators.
These institutions need to understand that preventing crime financing is a core mission and not merely a regulatory obligation. Doing it right is an incredibly compelling business case and enormously profitable opportunity that should not be done in isolation. It’s a systemic challenge in which banks should be cooperating more closely with each other (the Dutch announced a cross-bank collaboration last year – something that could be a wonderful catalyst of collective protection) and other vulnerable industries, notably hospitality, airline, and travel. The Canadians here lead the way with deep, cross-banking, airline, hotel, and entertainment coordination.
Crime causality that we’re organized to miss
AML, fraud, cyber, internal corruption, trade compliance, and trafficking are often different departments, with different budgets and different organizational focuses. More troublingly, often with different technologies. Why? Because that’s how budgets evolved and the specialized ecosystems that grow from those budgets. An evolution that left open a terrible set of flaws in our fight against financial crime.
We have found in the “Genome” project – more below – that complex, organized crime are rarely independent activities. The pathologies are all behaviorally related – money laundering and tax evasion are an obvious example. We have found in our work that when we find one crime, we know another is also happening, or we can predict it will happen through a causal network of relationships, transactions, and asset class alignments.
So in crime discovery, the sum of the parts is considerably more transparent than the parts themselves. This is particularly evident with the relationship between employees and crime. Simply put, if your analysis does not look for causality between different crimes, or at the very least you’re not coordinating effectively across departments, you’re missing more than 10% of exposure.
A coming change in regulatory energy and focus?
What of the regulators?
The past has seen a surprising lack of dynamism and agility. Though some regulators are becoming more aggressive and creative – the FCA should be called out here as a creative benchmark – most remain slow and academic in understanding the true crimes taking place.
The U.S. also saw some excitement in September 2018 when the four major regulators issued an incredibly encouraging note to the market stimulating innovation. In particular, the statement that “while the agencies may provide feedback, pilot programs in and of themselves should not subject banks to supervisory criticism, even if the pilot programs ultimately prove unsuccessful …… Likewise, pilot programs that expose gaps in a BSA/AML compliance program will not necessarily result in supervisory action with respect to that program.”
This was a massive opportunity for intellectual and entrepreneurial creativity, and some banks created “innovation labs” as a result. However, very little changed in the U.S. Examinations are still based on antiquated rules and behaviors from the early 2000s, using technologies that stem from the 1980s and 90’s bought not because of their value, abilities, or what they can catch, but because others bought them in the past and they map to what regulator’s felt were previously relevant two decades ago.
The EU’s 5th, and soon 6th Anti Money Laundering Directive is creating solid momentum. And in the US in 2021, we see renewed prospects of progress—from the US Congress, no less. On December 11, 2020, the Senate passed the Anti-Money Laundering Act of 2020 or Division F of the National Defense Authorization Act for the 2021 fiscal year. It’s way too long to summarize here, so I point to an excellent summary by Sullivan and Cromwell LLP and highlight four essential items:
- The establishment of the Bank Secrecy Act Advisory Group subcommittee on innovation
- An acknowledgment that current examinations and rules are outdated and need modernizing
- A streamlining of SAR filings (at last!!!)
- And “standards intended to allow for innovation, such as machine learning, with risk-based approaches to the testing and risk management of these innovative methods.”
Some would say it’s about time. Yes, but until now, the technology was not ready, and banks would just be wasting money. A lot of money has been wasted. Too many over-marketed startups, too many consultancy firms pretending to be innovators, too much hype from over-capitalized companies rebranding the same old technology. So why are things different now?
The AyasdiAI crime-fighting breakthrough
This is one of the few times that technology can be the catalyst for a win-win result. In all honesty, I don’t think it’s just Ayasdi making the difference. Several other firms also have the potential of market transformational change. But I’ll let you find out about them yourselves.
We’re working with some pioneering bankers that have said an absolute YES to that question. Not because of some regulatory scrutiny, but because they know they have to. This is not altruism in any way – they are bankers, after all. This is because they know speed is of the essence, that the market is changing, that examinations are on their way to changing, and so someone needs to lead the way.
Over the last year, we invested in projects with these firms. Not a lengthy PoC trying to deliver on a marketing promise, but a deep and honest discovery partnership, diving into the darkness that hides so much crime. These firms had spent $10’s of millions in big TMS deployments, have their own data scientist labs, and have been experimenting with some noisy, big marketing budget AI players.
Our projects were deliberately quiet. Partly because we weren’t sure it would work, and also because of the unusual angle we took, pulling oncological research (molecular characterization of tumors) into decoding massive amounts of data into a new multi-dimensional behavioral view with GraphML, TDA, and every other algorithmic approach we could muster into a single orchestrated discovery framework. We called this “Project X” or “Decoding the Financial Crime Genome” – to find key behavioral indicators or biomarkers that are often incredibly weak and ignored, that were true signs of criminal malignancy.
The objective was BIG:
- Fidelity: Discover 100% of crime within the institution’s data – laundering, evasion, fraud
- Efficiency and Productivity: Generate ZERO false positives
- Speed to Fidelity: Deploy in less than two months and use data already available
- Value Longevity: a fully dynamic framework so new crimes, asset classes, and regulatory change can be modeled and discovered in real-time.
It was fantastic fun to see all the orthogonal, out-of-the-box thinking. Many ideas failed, but that’s what innovation is all about – hypothesis, test, recalibrate, change the hypothesis and start again.
We didn’t hit these stupidly aspirational objectives. But we’ve got bloody close. From this lofty set of objectives, we can share the below results from a global bank currently using two horrendously expensive TMS:
- Fidelity: high confidence explainability, removed more than 80% of L1 and L2 alerts. L3’s (high probability of crime) and SARs increased by 30%, and we discovered a large (large with a $B) amount of new crime missed completely with current processes. All TMS high-risk alerts are now discovered before the old TMS found them. Some by over a year.
- Efficiency and Productivity: The false positive-to-L3 ratio went from 45:1 (old TMS) to 3:1 and an overall increase of more than 500% in FIU productivity
- Speed to Fidelity: From pre-deployment to value discovery in 2.5 months using the same data as the current TMS used. So best to say 6 months accommodating internal bank processes and timelines.
- Value Longevity: Fully dynamic optimization of current TMS rules, integration of new behavioral discovery mechanisms, and full regulatory reconciliation all in place
- And for the cynics out there – all governed by a contract-based SLA, so no bait and switch.
Douglas Adams’ “Hitchhiker’s Guide to the Galaxy” has on its back cover – “Don’t Panic.” Innovation in a regulated market absolutely needs that sentiment and confidence. So, don’t panic. We’ve got it covered. It is not a silver bullet, but we’ve got a solution focused on not just true, real, material transparency, but with cost efficiency and explainable effectiveness as a foundation to a revitalized AML business unit: frictionless deployment, rules-like granularity in explanations and justification of results, and wholesale reuse of the investments made in the past. It is time for a comprehensive upgrade to catch up with a market that has outmaneuvered the AML function. Symphony AyasdiAI is doing just that.
We spent the last year testing our technology and approach. I’m proud as hell to announce it as a commercially available solution.
Sneak peak, we’re going to call our new crime-fighting solution Ayasdi Sensa-NetReveal. It’s an elegant, easy-to-use application suite that enables top-down (CRO, CCO, and CAMLO) and bottom-up (Investigator) analysis and collaboration across every crime category. We intended to crack the code on laundering and fraud, and we succeeded. No more multi-year deployments thanks to a solution that is up and running in weeks (another SLA), dynamic and evolutionary in coverage and functionality, and perfectly facilitates a regulatory audit.
The acknowledgment of a moral challenge, and what’s next?
After a decade of over-promised, failed technology promises, I’m sure the cynics are skeptical of our confidence. We get that. Others may promise similar results as they desperately grab PoCs to deliver on exaggerated sales assurances. Or the understandable concerns about the mess of data hindering a decision to move forward. We’ve got that covered as well, though we will not discuss that in a public forum.
There is no doubt that technology often makes over-excited commitments that end up stumbling. But every now and then, there is a leap. We’ve spent a year ideating and innovating, testing, failing, and ultimately succeeding. The Symphony AyasdiAI approach is groundbreaking, and the confidence we have in what this team has pioneered is extraordinary, and such innovation will always be viewed with skepticism. But, as an anti-money laundering team, you’ve got a unique opportunity to partner with AyasdiAI and what we have achieved.
So, it is up to you—the CAMLO, BSA officer, and CCO—to decide whether you want to truly solve these challenges, or at the very least, make a real impact against adversaries. Fortunately, a number of you see the value in the technology and are boldly testing the regulatory direction, with the understanding that you are leading the pack. That, combined with new incentives from the regulators, and we may all be back in the race.
Or we continue the ineffectual orthodoxy institutionalizes the very crimes we seek to prevent with outdated, expensive, and increasingly failed solutions that have missed criminal activity for more than a decade. It’s time to change and solve this problem. Or at least, catch up to the sophistication of modern crime.
Enough already. Let’s get going.