04
Jan 19
Compliance

AI stakes claim for supporting role

Banks are beginning to embed advanced technologies at the heart of financial crime compliance

The Sibos 2017 compliance stream in Toronto made a clear case for the deployment of advanced technology and greater collaboration with law enforcement agencies to improve the efficiency and effectiveness of banks’ financial crime compliance efforts. In Sydney, the compliance sessions at Sibos 2018 revealed that encouraging progress is being made on multiple fronts, suggesting good reason for a positive outlook.

Taking sanctions screening as an example, HSBC global head of correspondent banking Barbara Patow recognised that bank processes are not always as efficient as they could be, and staff resources can be drained very quickly without adequate support from technology. But screening has become much more effective in recent years, she said, and HSBC has seen positive results from the use of artificial intelligence (AI).

“We’re now in the phase where we are getting more efficient and there is greater awareness with regard to how we can actually do this,” said Patow. “At HSBC, we screen 1.2 million payments a day – we stop 75,000 payments a day and we are now applying machine learning and AI to about 35,000 of those alerts. We still have human intervention to work those final alerts, but that’s a huge step forward to ensure we can meet demand in real time.”

Accelerating momentum

Many banks are still in the testing phase with new technologies, but momentum has clearly accelerated, thanks in part to the need to alleviate pressure on human resources involved in transaction monitoring and sanctions screening. Banks and other institutions are demonstrating increased enthusiasm for AI and robotics to tackle financial crime, panellists observed, noting also that such technologies can be bolted onto existing systems, providing a fix to incumbent solutions.

“Most of the industry is working on use of machine learning as an additional filter to existing rules-based systems that are producing false positives. The question is whether one can use enhanced analytics and machine learning to look at the results and enhance the quality before you put them to your analysts,” said Adrien Delle-Case, policy advisor for digital finance regulation and policy at the Institute of International Finance (IIF).

A survey of 59 financial institutions (largely banks) published by the IIF prior to Sibos 2018 revealed that 69% of respondents are already using or experimenting with machine learning techniques in their AML-related analysis, while only 2% have no plans to do so. Machine learning will not fundamentally change AML but rather it should lead firms to enhance and rethink their processes, the report suggested.

“Financial institutions need to communicate so that staff understand they will not be replaced by machines. These people are trained; they are experts in their field and they know how to recognise suspicious activity. The aim is to empower them and refocus resources on problematic cases and risk assessments,” said Delle-Case.

Encouragement also came from Miles Ward, global head of solutions for Google Cloud, who said that while the challenges in banking might seem insurmountable, other sectors actually operate on a much larger scale with even greater volumes of data. Nevertheless, higher levels of automation did not equate to obsolescence, he reiterated.

“Tool after tool is being built not just to simplify the automated work of identifying bad actors and criminal activity, but also to improve the work experience of every one of the professionals trying to solve those problems. Some of this is about replacing people, but it is also about augmenting people and giving them superpowers,” said Ward.

In the near term, banks must continue to test new technologies to determine their merits as well as exploring the extent to which human intervention will still be required for efficient, effective compliance. Handing over complete responsibility and autonomy to machines might present stiff challenges in the area of financial crime, given humans can very often disagree on what constitutes suspicious activity.

“Many global financial institutions are spending more than US$1 billion on financial crime compliance. We can’t continue to put bodies against the enormous amount of data we have coming in, so we need to find solutions. One of the challenges here is that reasonable people can disagree on the outcome of a scenario. There are a lot of grey areas,” said Michelle Neufeld, head of compliance and operational risk for Wells Fargo’s financial institutions group.

Real-time compliance

Meanwhile, transaction banking as a whole is embracing new technologies and faster, more automated processes. With the current proliferation of new payment channels, the rise of cryptocurrencies and the move towards real-time payments around the world, the challenge of dealing with financial crime shows little sign of diminishing in the near future. If anything, banks will need to be even faster and more innovative in meeting their compliance obligations if they are to maintain resources and budget at a manageable level.

Natalie Hall, general manager for financial crimes compliance at Commonwealth Bank of Australia, emphasised the need to know your customer’s behaviour, rather than concentrating only on standard know-your-customer routines.

“We want to understand our clients’ policies and programmes so that we can look at our output from transaction monitoring and sanctions screening and check that it aligns with what the client told us about their AML programme and their risk profile. The difficulty for banks is to then verify that information in a meaningful way and demonstrate to our regulators that we truly understand the risk,” said Hall. 

Collaboration between law enforcement agencies and financial institutions in the fight against financial crime is far from a new concept, but the recent wave of public-private partnerships (PPPs) highlights the progress that has been made in sharing information and breaking down barriers to catch the bad actors.

James Freis, chief compliance officer at Deutsche Börse Group, referred to a PPP 2.0 model; a more structured way of sharing information with greater resources than the informal entities that existed a decade ago. Such groups operated in a more ad hoc and low profile way than the PPPs that have been established in recent years.

“The difference now is the resourcing and the structure,” said Freis. “We have dedicated resources and better people, there are 10 times the compliance staff we had a few years ago, and people with different backgrounds have been brought in to take this to a higher level. We also have better access to data than ever before, we can look at data in a different way, and in certain jurisdictions we have a clearer legal basis to share information.”

Following on the heels of the UK’s Joint Money Laundering Intelligence Taskforce (JMLIT) and the development of PPPs in Canada, the US, Hong Kong and Singapore, Australia formally launched its own PPP, Fintel Alliance, to combat money laundering and terrorist financing in March 2017.

With the backing of AUSTRAC, the Australian government’s financial intelligence agency, and participation from the top four Australian banks as well as HSBC, Paypal, Western Union and other official agencies, Fintel Alliance has been lauded for physically seconding experts from the public and private sectors, rather than simply meeting intermittently to share information. A forthcoming SWIFT Institute report will explore Fintel Alliance’s role and structure in more detail and provide guidance to jurisdictions planning to establish their own PPPs.

“Fintel Alliance is quite unique,” said Simon Norton, analyst in the strategic policing and law enforcement program at the Australian Strategic Policy Institute, and author of the SWIFT Institute paper. “In other PPPs, people essentially get together for meetings and then go back to their usual roles, but we have co-location of professionals, which is really important.”

Norton believes the key criteria for an effective PPP include voluntary participation – his report will feature organisations that chose not to be involved and explain their reasoning – and proper resourcing. In a highly-concentrated banking market, the participation of Australia’s four largest banking groups has also been critical to Fintel Alliance’s success, Norton added.

While there are inevitable legal and operational issues that need to be tackled to make a PPP work successfully, the biggest challenge may be cultural. “We still have a cultural issue which is the risk and unintended consequences of information sharing – both in the private sector and in law enforcement, some are more reticent than others and this is something PPPs have to work through,” said Norton.

Evidence is already mounting on the wider benefits of information sharing between the private and public spheres. During a session dedicated to PPPs, for example, details were revealed of how Fintel Alliance participant Western Union had raised concerns about small credit card transactions that it suspected were linked to child exploitation. By pooling data and intelligence among participants, the PPP identified previously unknown offenders and made more than 20 referrals to law enforcement agencies.

“There was a 316% increase in child exploitation-related suspicious matter reports and suspicious activity reports over 12 months, but critically it advanced awareness of the issue within the financial services industry and it got people talking, because no one wants to see children exploited,” said Paula Chadderton, international counter-terrorism adviser at the Australian Department of Home Affairs’ Centre for Counter-Terrorism Coordination.
 

Technology
December 19, 2018
Innovation
January 9, 2019