Introduction:
Automation in the regulatory space has immense value as regulations are constantly being updated or changed, and it can be hard for risk or compliance teams to keep up. AI-powered platforms are rapidly becoming a staple in the RegTech space, but they need to be trained by the current experts – humans – to attain the desired precision rates.
The Client:
We were approached by a regulatory change management (RCM) solutions provider and a leading innovator in the AI-based RegTech space for our expertise in training AI tools.
The company’s AI-powered platform automates and streamlines the tracking of compliance activities in the financial sector as and when regulatory changes happen.
This allows chief compliance officers and general counsel in the financial services sector to stay on top of any changes that might affect their business and respond to the evolving regulatory needs to remain compliant.
The Challenge
Financial regulations need to be continuously monitored so companies can avoid breaches and enforcement fines. This makes real-time, automated regulatory updates a very powerful feature for compliance teams.
AI-powered platforms in this space need to be trained to track regulatory updates with high accuracy.
Cenza’s task for the client was to train the ML algorithm
- to categorize news articles and financial updates into a pre-defined number of topics
- automatically extract enforcement actions, including violations and penalties
- automatically extract obligations, constraints, and rights
The Solution
Human in the Loop(HITL)
Cenza’s primary goal while training any AI platform is to ensure that the platform consistently and efficiently does the task assigned to it.
To this end, we utilized a human-in-the-loop strategy, which comprised deploying a team of expert compliance analysts to work with the client to fine-tune the platform’s performance, while also flagging the platform’s errors.
The RCM solutions provider’s AI platform is designed for document classification: it classifies, tags, and labels financial blogs, updates, news articles, regulatory documents, whitepapers, and so on, into topic-specific categories.
To ensure complete accuracy in the classification process, the platform needs human expertise. Our team of compliance analysts validates the output from the platform and helps refine the platform’s document classification feature. This method of supervised learning is teaching the tool to better identify the categories, thereby offering greater accuracy to the categorization. In addition, we are able to scale up quickly when the need arises.
Accuracy is critical for document classification and Cenza’s robust quality control standards are ISO 9001 certified.
Annotation
In addition to training the tool on classification, our team of compliance analysts helped the client in annotating and extracting:
- enforcement actions such as violations and penalties from case files, litigation documents, and regulatory updates
- obligations from financial updates, notices, and circulars from various regulators such as the SEC, FINRA, FinCen, and FCPA, among others
Cenza takes client data security very seriously, and has implemented ISO 27001 controls, state-of-the-art infrastructure, and is aligned with GDPR norms.
The Impact
- Within a span of five months of incorporating the Human in the Loop (HITL) process, the tool had processed over 20,000 documents and had extracted 90,000 attributes or concepts.
- Cenza’s human review process has helped improve the AI platform’s accuracy in terms of automated identification of various attributes from multiple publications, document types, and geographies.
What are your AI project delivery challenges?
We can help.
