If regulatory reporting became machine-readable, it would reduce friction and could lead to more profound benefits. Nicholas Pratt reports.
One of the biggest drawbacks in regulating the financial services sector is that supervisors have so few resources in comparison with the firms they are supervising. So it is understandable that regulators are turning to new technology.
The UK’s Financial Conduct Authority (FCA) has been at the forefront of this trend, not least in a bid to develop a rulebook that can be machine-read. In June 2017, Nick Cook, the FCA head of data and information operations, told the audience at London Fintech Week that the aim was to “put out rules which are written manually in ways that can be fully and unambiguously interpreted by machines”.
A few months later, the FCA and Bank of England held a two-week ‘TechSprint’ on regulatory reporting. The workshop gathered financial services firms, technology vendors and compliance experts to explore the potential for model-driven, machine-readable and executable regulation.
Then in February 2018, the FCA issued a ‘call for input’ to further explore the idea of a reporting regime that could be automated. Reaction from providers of compliance services has been overwhelmingly welcoming.
“Changes in compliance are being driven by regulators, not firms,” says Jeanette Turner, chief regulatory officer at regtech firm CSS.
“Regulators with limited resources are turning to technology and data to spot trends, identify risks, and to better target firms for examinations. Regulators now want to receive data in structured data formats, changing the make-up of compliance departments at firms.”
Firms must have a holistic view of data across departments, various systems, and across borders, says Turner, and machine-readable rules will make this much easier to achieve, and solve the ‘garbage in, garbage out’ problem in reporting.
“With machine-readable rules, regulators state exactly the data they require and systems on the other end provide that data which requires no extra work for the firms. Regulators can also change the rules with little notice, and quickly receive the new data required,” says Turner. Not all reporting requirements can be translated, but it is a great start, she adds.
David Pagliaro, regional head of State Street Global Exchange, also welcomes it and illustrates the resources needed to support reporting as it stands today.
“We have 60 staff currently working on constantly evolving rules. One of our business partners has another 180 staff worldwide doing the same thing,” he says.
Four steps
For asset managers, there are generally four steps to compliance, says Pagliaro.
Firstly, be aware of the rule; secondly, working with a range of service providers, try to understand it; thirdly, understand if compliance will involve changes to governance or operating models; and finally, make the necessary system changes or upgrades.
Machine-executable rules “really slims down” this end-to-end process, he says.
To make machine-readable rules a practical reality, there will need to be more standardisation and less of the principles-led ambiguity that sometimes obstructs the compliance process, Pagliaro says.
Established rules like Ucits have very well-understood definitions around aspects – such as the allowable levels of transferable securities – whereas new rules tend to lack clarity in their early stages. However, if future regulations are designed with machine-readable processing in mind from the outset, it could make for a much clearer and more standardised regulatory regime.
“Many of the rules that exist today come from guiding principles but this approach is not relevant in the digital age,” says Pagliaro.
Another discussion area is the introduction of blockchain in compliance. “Instead of an asset manager sending 100 text files every day, we could move to a scenario within a secured network where all reports are imprinted and all counterparties are included. I see the two technologies [machine-readable compliance and blockchain] as complementary.”
Machine-readable regulations would be very valuable in driving automation, says Andrew Yuille, head of risk business solutions at Thomson Reuters. “We already monitor changes in regulation and wrap these with taxonomy and metadata to automate regulatory change management, so making the original regulation machine-readable will simplify this work.”
But to be truly valuable, regulators across all sectors and countries would need to adopt the same taxonomy, he says, adding that there could be greater value from this approach than the simple but important task of driving automation.
Thomson Reuters has worked on monitoring changes in a regulation and on capturing enforcement data relating to each paragraph of a regulation, and Yuille says it is easy to see how this approach can evolve to the point where artificial intelligence (AI) is used to automate the creation of trading or business controls, audit plans and other compliance measures.
Profounder benefits?
Lombard Risk was one of the technology firms that participated in the TechSprint. James Phillips, its global head of regulatory strategy, expects progress to be swift. “Machine-readable rules will happen and will happen fast,” he says.
Moving to a machine-executable compliance world, and codifying rules where possible, leads to a question of whether the benefits would go beyond automation and processing efficiency to something perhaps more profound. For example, could the predictive abilities of AI and machine-learning technology be applied?
“If machine-readable regulation led to a more granular style of reporting, the regulator would be able to rely on more information that was standardised and unambiguous,” says Phillips. “Consequently the regulator would have a data set that can be reviewed with AI-based tools and then analysed information can be fed back to the regulated [firms].”
The industry’s work on machine-readable rules could be a preliminary step on the path to machine learning, says Phillips, while also making clear that the two technologies are very distinct.
“We could see a new regulatory regime embodying continuous process improvement. If you were able to apply machine learning to the process and generate information that could be fed back into the system, it could turn the compliance and reporting process into a virtuous circle,” he says.
Such a development would create a redefined data relationship between the regulator and the regulated where reporting is based in the cloud and “always on”. As the use of machine learning becomes more practical, the industry could even look to develop AI-based central utilities operating for the benefit of the regulator and the regulated.
“There are classic-model AI services out there that buy data from private sources, apply some AI learning and then sell that enhanced information on to its subscribers. What we could see in the future is industry-wide, forward-looking, pooled data informed by federated private machine learning that would be operated not for commercial application, but as a common utility,” Phillips says.
There are numerous challenges, though – some of them technical, many of them commercial, he adds – and it would require everything to sit in the cloud, yet be highly secure. Then there are questions of governance, ownership, maintenance and so on.
But should these challenges be overcome and the adoption of AI and blockchain tools continue to accelerate, the industry could yet find that it is possible to provide a predictive compliance outcome for both regulator and regulated, says Phillips.
Other providers in the regtech space see the development of machine-readable rules as a welcome disruption to the interpretation of legal documents that will complement compliance services looking to apply behavioural science and cognitive computing to the reporting and surveillance process.
“In the future, it could lead to certain regulatory reporting feeds acting as an input into our system,” says Taras Chaban, head of buy-side solutions at Nasdaq. It may be a work in progress, says Chaban, but for other AI technologies to really flourish in the compliance world, it does need more machine-readable data that can be algorithmically processed.
“Ultimately it is about simplifying the process and making compliance officers more efficient and with a better understanding of a portfolio manager’s decision-making. We are seeing the same thing happening in other industries, so it is inevitable that the funds industry will evolve in the same way.”
©2018 funds europe
