IBM Digital KYC on AWS uses Generative AI to transform Client Onboarding and KYC Operations

By: Nisha Dekhtawala, Sr. Partner Solutions Architect – AWS
By:Diego Colombatto, Prin. Partner Solutions Architect – AWS
By:Rick Hoehne, Sr. Partner, Financial Crime – IBM
By:Abhay Mhatre, Associate Partner, Data & Technology Transformation – IBM
By:Sanal Kumar, Partner Developmenet Manager – AWS

Introduction

In today’s digital world, institutions face significant challenges with onboarding new customers and managing their risk through the lifecycle of their relationship with the Institution. These include adding friction to the customer experience, retaining highly skilled and expensive analysts, dealing with rapidly evolving regulations, a far more complex sanctions challenge, AI deepfakes, increased financial liability when criminal elements are onboarded, and legacy platforms that are not able to keep up with these evolving requirements.

Conducting customer due diligence and enhanced due diligence throughout the lifecycle of the customer’s relationship is labor-intensive, requiring highly skilled analysts that are expensive to train and hard to retain in today’s labour market. Many institutions are finding this to be an “unsustainable” approach to assessing risk and meeting Know Your Customer (KYC) regulatory obligations. In addition, the customer experience is becoming more and more frustrating to new clients, resulting in abandonment by new customers opening accounts. Regulators are also adding to the challenge by shifting focus from rules to outcomes and putting increased pressure on finding deeply hidden sanctioned entities. As institutions move from 100% periodic reassessments to providing continuous reassessments, this too is straining existing capabilities.

Celent estimates spending on Financial Crime (including KYC) technology and operations by financial institutions worldwide in 2023 was US$ 58.0 billion (and it continues to rise). Institutions that still rely on human experts operating manual KYC processes, with data that is dispersed across multiple systems are unable to keep up. For large institutions, multiple KYC solutions across their landscape further exacerbate the problem by creating technical debt and putting more pressure on operations to learn multiple systems.

By working with IBM ecosystem partners, IBM’s Digital KYC offering provides an alternative by introducing a flexible, API first platform that allows for the creation of AI-generated digital workers to automate many of the labor-intensive tasks. It also provides a dynamic workflow and case management capability to radically transform the approach to meeting today’s evolving regulatory obligations and expected customer expectations.

IBM Digital KYC Solution

Platform-Based Approach – Reducing Technical Debt with a single Enterprise platform

Digital KYC employs a modern approach that transitions from standalone departmental solutions to a common configurable and dynamic platform that can be adapted to each division’s unique requirements. This simplifies the technology landscape while accommodating localized compliance requirements via a dynamic platform with common workflow, orchestration, integration, and data foundations. By providing the offering “as a service” from the AWS cloud, the offering also reduces the cost of enterprise infrastructure that supports the future growth and expansion of the institution.

Document Centric Approach – Reducing Customer Friction

The most common frustration during the onboarding process is asking the client for information the customer thinks you already know, or re-asking for information if the institution realizes they incorrectly classified the risk. Digital KYC radically transforms the customer experience by using Generative AI (Gen AI) to read Corporate Policy Documents, identify and request the needed evidentiary documents, and then extract needed data from these documents. Ideally, when opening an account, the customer is only asked for the bare minimum amount of data required to meet the “right” corporate policy that could not be found in the document, enterprise data, or trusted 3rd party data sources.

This intelligent, document-first approach reduces customer friction, accelerating onboarding from weeks/months to days, and reducing the amount of time spent originating and tracking requests for information.

Increased Automation – using “Digital Workers” to improve productivity and lower costs

Digital KYC’s vision is built on a data-driven operation where “digital workers” to automate labor-intensive tasks that require specific expertise, that can be integrated with our dynamic workflows to dramatically improve the customer experience while reducing costs. For example, in Digital KYC, we can read policies and customer submitted evidence documents to map collected data with policy obligations. Further, a digital worker can request additional client information and track responses. Therefore, the analysts’ role changes from a “maker” of KYC due diligence information to a “checker” of information collected and curated by the digital workers. This approach also minimizes the level of effort required to implement minor policy changes.

By using digital workers to monitor transactions and data sources, we can also automate KYC refresh cycles by assessing the impact of any material change in the risk profile of the customer, and triggering an automated refresh of KYC data.

The end-to-end operational model consists of seven areas of automation and AI:

• Policy-based Customer Due Diligence (CDD) and Enhanced Due Diligence (EDD) process orchestration,
• Automated collection of evidentiary documents and data to meet policy and compliance obligations,
• Context-aware extraction of required data from unstructured and complex documents,
• Gen AI assisted risk assessment of complex network relationships,
• Curation of relevant policies and collected documents and data into pro forma narratives,
• Material changes event monitoring to enable continuous risk monitoring and assessments,
• End to end dynamic workflow including governance, monitoring, and tracking.

Figure 1: IBM Digital KYC high-level end-to-end process

1. Customer uploads documents: The customer provides the necessary documents for account opening to begin the process.
2. Data extraction and analysis: Gen AI reads the bank’s policy documents and determines the documents and data required to meet policy obligations. Gen AI then reads the documents submitted and extracts the relevant data. Gen AI also accesses the internal master data and 3rd party data to determine if complete data is available.
3. Missing data request: If data is missing, the system asks the customer to provide it. This is an iterative process – the system repeatedly requests additional information from the customer until all policy requirements are met.
4. Narrative generation: Once the system meets the policy obligations, it creates a narrative and sends it to the analyst.
5. Analyst approval: The analyst reviews the pro-forma narrative and approves it.
6. Onboarding completion: Upon approval, the system notifies the customer, and the onboarding process completes.
7. Continuous monitoring: The system continues to monitor customer data for changes that might affect their risk rating and initiates updates as needed.

Throughout the onboarding journey, the platform continuously assesses the next best action based on triggering events. These could include completing the onboarding or refresh process, triggering outbound requests for information, or involving a human analyst when needed to resolve issues.

Leveraging Gen AI

IBM’s elegant approach of using Artificial Intelligence for KYC combines targeted innovation with rigorous model governance. As described earlier, a highly manual and time-consuming aspect of current KYC processes is reading and extracting critical data elements from complex documents, comparing them against policies, and summarizing the relevant information to perform risk assessments and generate customer narratives. Today, KYC analysts must analyze extensive and intricate documents like articles of incorporation, trust documents, and regulatory filings. Analysts can often spend up to 1 full hour reviewing certain types of documents. Analysts must also chart out complex relationships to identify false fronts used by sanctioned entities.

Digital KYC uses Gen AI to create “digital workers” to assist human workers by reading and curating relevant policy and collected information during the due diligence process. This addresses the “unsustainable” situation faced by today’s institutions by assisting the analyst in finding and organizing needed information. While the process itself is highly automated, only a trained analyst can review and confirm the information curated by the offerings digital workers, ensuring that all regulatory obligations associated with the use of AI are met.

Figure 2: Document data extracted, compared to policy and summarized for Next Best Action

Digital KYC also helps an analyst quickly uncover complex relationships that sanctioned entities use to mask their intentions. By layering a Gen AI natural language interface atop the graph database, it provides a human-readable analysis to support KYC risk assessments. This simplifies uncovering beneficiary counts, customer geographic interactions, and high-risk entity relationships during risk assessments. This minimizes the financial risk associated with unintentionally onboarding sanctioned entities or other criminal elements, while reducing the time the analyst spends in completing such analysis.

Figure 3: Complex relationship information automatically extracted and translated in human readable output, using IBM Digital KYC Gen AI capabilities.

Digital KYC Technical features

At the core, an operational workflow application will continue to be used for orchestrating the overall business process. However, productivity will be increased significantly by automating manual tasks using a set of AWS Services: 1) using Amazon Textract to extract data from all submitted documents, 2) using Amazon Bedrock to automate comparison of extracted data to business policies and suggesting next best actions as well as deriving entity risk.

Digital KYC uses AWS serverless functions to automatically scale-up and scale-down resources to follow application load, further optimizing costs and performance. Different Amazon S3 storage classes are used to further reduce costs, depending on document access frequency.

Security and Compliance are key concerns for any business. AWS Key Management Service (KMS) is used to manage keys and encryption of all customer data, enabling to achieve the required level of privacy, security, and compliance. External keys can be used, in case customer prefer to keep cryptographic material outside AWS, such as on-premise. Integration with AWS CloudHSM or third-party HSM is also possible. AWS CloudTrail is used to provide detailed audit trails of all user and system actions, critical for supporting regulatory audits and demonstrating compliance. AWS Identity and Access Management (IAM) is used to implement granular control over access to data and resources, with support for multifactor authentication. Amazon Certificate Manager provide secure management of X.509 certificates for SSL/TLS connections, securing data in transit. AWS Secrets Manager centralizes and secure the secrets required by the application, such as API keys and data repository credentials.

For automation and performance, Amazon CloudWatch monitors resources and application in real-time, enabling automated actions as required, while AWS CloudFormation automates provisioning, configuration and re-configuration of resources.

Conclusion

This blog highlighted the challenges financial services customers face in meeting KYC regulatory requirements. Digital KYC can reduce onboarding and compliance costs and reduce turnaround time by automating expensive labor tasks. It also improves ongoing client lifecycle management by automating file refreshes triggered by events that change risk assumptions. We estimate that Digital KYC can reduce workloads by an order of magnitude compared to manual processes, while simultaneously improving the financial institution’s response to regulators and improving their customers’ experience by reducing the traditional “back and forth” of collecting documents and answering questions.

In the future, we expect the KYC functions will be more integrated with the institution’s counter fraud, security, and other risk assessment functions. For institutions with modernized counter fraud programs, the KYC platform will integrate with the counter fraud program to monitor transaction profiles and identify changes to risk profiles, while KYC feeds customer risk and vulnerability scores into fraud decisioning. The strategy prompts policy and strategy changes in a timely manner, providing a unique opportunity to keep the institution secure at lower costs with improved customer experiences.

IBM Digital KYC can be operational within weeks for a subset of customers/products and can evolve across the enterprise. IBM Consulting can run a proof of concept to validate sample KYC data files, ensuring completeness, consistency, and currency. Contact your IBM or AWS client representative to learn how IBM Digital KYC can benefit your organization.

.

IBM – AWS Partner Spotlight

IBM is an AWS Premier Tier Services Partner with numerous AWS Competencies and Service Delivery specializations, including Migration Consulting competency and Generative AI Services competency, IBM holds a distinguished position in the industry. With IBM Consulting’s extensive expertise and depth of capabilities in Cloud adoption gained through thousands of client transformation, businesses trust in IBM’s proven track record of success.

Engage with IBM Consulting for a proof of concept (PoC) or demo.

Contact IBM | Partner Overview | AWS Marketplace