The financial services industry is facing stricter regulations than ever before. As the ability to collect and harness the power of data grows, so will compliance requirements from regulatory bodies. While this keeps businesses and their customers safe, it is a considerable burden for any organisation to stay comprehensively compliant.
This is where regtech, or regulation technology, comes in. It uses AI to help businesses automate and manage their compliance programs. Instead of compliance officers being forced to pour through thousands of paragraphs of regulatory text, this information is transformed into calendarized workflows. It can also provide valuable insight with curated reference documents and alert customers when important regulations change. When this comprehensive and automated delivery of intelligence is integrated into firms, it not only eases the process of compliance but also enhances their agility with future risk.
This is all made possible by teams of data scientists who are building advanced technology and scripts that:
- Automate specific steps in the process.
- Guide human action within the process.
- Ensure quality assurance by surfacing and prioritising any potential errors and anomalies.
The RegTech Process
Regtech has a number of potential outputs and end products, including digital identification, more secure two-factor authentication, and the standardisation of compliance formalities. The most complicated one, though, is workflows.
Workflows break down all regulatory rules and supporting information into functional and calendarized tasks.
This is a multi-step process where scripts analyse all of the regulatory text on regulators’ websites. The scripts then identify the information type, categorising and separating them between regulations or rules and supporting information, such as definitions and further clarification.
The scripts then extract the metadata from the website. This includes information ranging from the document title to the date published. This allows regtech users to search the content for keywords easily.
Finally, the scripts pinpoint where and how the regulatory information applies. For instance, when the website publishes regulatory content with various definitions, the scripts identify where they should be employed within the overall obligation. Then, the workflow creates a customer map of the information for each user—this map can be based on the logistics, processes, and systems within each user’s organisation.
In the end, scripts allow users to log into their regtech platform and see a dashboard customised for their needs—every piece of information and each workflow is explicitly tailored to their business.
The RegTech Approach To Risk
Failure to comply with industry regulations can be fatally detrimental to any company, both in terms of fines and reputation damage. Regtech firms consider this risk and understand that inaccurately interpreting content from regulators is often what causes these compliance shortcomings. To adequately address this risk, regtech developers involve humans in the process of deciphering the regulatory text. They rely on regulatory intelligence experts and have them work alongside AI to distinguish between actual regulatory obligation and supporting documentation. This human-AI collaboration will likely always be necessary to enhance accuracy and reduce risk.
However, the risk is not limited to just the step of interpreting regulatory text. Each step of the compliance process comes with its level of risk. Regtech firms measure these risks and, depending on the risk level, design an appropriate collaboration between human experts and technology.
Teaching RegTech Models
Regtech data scientists are continuously engineering their models, determining if any new input data should be added to optimise their output. To accomplish this, they run the model and compare its output against their predictions, which is also known as error analysis. From the results of this analysis, data scientists can complete feature engineering—a fundamental process in machine learning that identifies whether models need to know more information. Then they run the model again and again, cycling it through feature engineering, modelling, error analysis, and output. Over time, this continuous process enables regtech models to learn more and more until they can outperform humans.
The RegTech Result
When it comes down to it, though, businesses need to know if the process works. This can be seen by how long it takes an enterprise to analyse new regulation and adjust processes to comply. Businesses will comb through all of their regulatory text and curate it themselves. Manually, this can take hundreds or thousands of hours. Regtech analyses and curates the same regulatory text in a matter of minutes—it is also more accurate.
AI And Humans
Regtech developers do not create their models in a vacuum—or a silo, for that matter. Throughout development, there is an exhaustive collaboration between regulatory intelligence professionals and data scientists, a relationship which is crucial. This side-by-side effort enables regtech developers to translate expertise in regulation and law into AI models.
Often, regtech development works in much the same way as Ford’s first assembly line. While initially Ford’s process was comprised entirely of humans, whom each performed a specific task, the humans were gradually replaced by machines. Regtech is taking what once was a wholly human process that took thousands of hours and identifying areas where humans can be removed. It replaces them with machines or scripts that probe through regulatory text and extract meaningfully, actionable information in a few minutes.
Regtech’s goal is not to entirely replace humans, but rather to shift their role. Technology will be able to take care of much of the manual, rote work, allowing humans to pursue the high-value tasks that only they can do.
Regtech values simplicity. It uses the most basic model to solve a problem, benchmark it, and then build out the model’s capabilities. To accomplish this, it strives to fully and deeply understand the business problem in front of it. Regtech firms work with their business customers to create models that address the issue. So while they rely on advanced technology, their end goal is not to use technology, but rather to serve the business objective—it is this goal that makes technology useful and not just the technology in and of itself.
Then there is the ethical aspect of what regtech does. AI will always have ethical implications. A prime example of this is Microsoft’s AI, which learned racism from Twitter. When AI is created and ‘taught’ by humans, there is the risk that it will adopt their biases. Businesses that use regtech will need to be assured of the integrity of the technology–that it will uphold the standards of their business processes and culture. Additionally, ethical compliance is a fundamental benchmark that companies in every industry strive to attain. The goal is not merely meeting basic regulation, but meeting regulations in the ethical way that the public and customers expect. So it is these implications that regtech firms take into consideration while they are building.
To find out more about the philosophy and technology behind regtech, please contact us.