Utilizing Generative AI in Southeast Asian Financial Institutions: Key Considerations for Implementation

Insight

Oct 17, 2024

Insurance
Banking/Capital Markets
AI

Generative AI is a transformative technology capable of creating new data and content. With its ability to comprehend and generate human-like text, it is rapidly being adopted by businesses worldwide to enhance operational efficiency.

In Southeast Asia, where rising labor costs have become a significant concern, reducing these expenses through process optimization is a critical priority. The financial sector, including banking and insurance, has long struggled to automate complex tasks such as underwriting, contract management, and investigations—areas traditionally resistant to machine-based solutions. Generative AI presents a promising opportunity to streamline these processes more effectively.

In practice, leading banks in Singapore have already begun integrating generative AI across a variety of functions. These include internal document search capabilities, transcription and analysis of customer interactions in contact centers, as well as more basic tasks such as content generation, research, and ideation.

This paper outlines real-world case studies gathered from client engagements, highlights key considerations for successfully implementing generative AI, and introduces ABeam Consulting’s suite of solutions tailored to financial institutions in Southeast Asia seeking to leverage this innovative technology.

About the Author

Atsuyuki Mori

Director Head of Insurance sector in Southeast Asia
Eiji Matsumoto

Senior Manager
Kentaro Yagi

Senior Consultant

Contents

1. Case Studies: Generative AI Applications in Southeast Asian Financial Institutions
2. Key Considerations for Implementing Generative AI in Southeast Asian Financial Institutions
3. ABeam Consulting’s Generative AI Solutions

1. Case Studies: Generative AI Applications in Southeast Asian Financial Institutions

Generative AI is being applied in a wide range of functions across financial institutions, though its usage often remains focused on more generic tasks such as text generation and ideation. Many financial institutions have yet to extend its application to more complex operations that require deep industry expertise. Historically, these processes were deemed too intricate to automate, as they involved significant human oversight and incurred high recruitment and training costs. The potential for substantial efficiency gains in these areas is therefore considerable.

Figure 1 presents real-world case studies of generative AI usage, derived from our consultations with various banks and insurance companies across Southeast Asia.

One notable example in the banking sector is the application of generative AI to streamline anti-money laundering (AML) operations. Tasks such as new customer investigations as part of Know Your Customer (KYC) processes, as well as fraud detection in suspicious transactions, typically involve sifting through large volumes of data, including online resources, regulatory guidelines, and historical reports. By leveraging generative AI, these labor-intensive processes can be significantly optimized.

In the insurance industry, there is an increasing need to apply generative AI to high-level decision-making tasks, which have previously been resistant to automation. For example, automating damage assessments following vehicle accidents and enhancing underwriting processes are areas where generative AI can drive considerable improvements. Additionally, there is growing demand for AI-powered solutions that support decision-making, such as automating the drafting and revision of complex documents, including contracts, risk assessments, and regulatory reports—tasks that traditionally require deep operational and expert knowledge.

Figure 1. Case Studies of Generative AI in Southeast Asian Financial Institutions

2. Key Considerations for Implementing Generative AI in Southeast Asian Financial Institutions

Successfully improving business processes with generative AI requires more than simply deploying new systems. Companies must begin these initiatives by thoroughly analyzing the specific operations they aim to enhance, clearly defining the challenges they intend to address, and outlining the desired outcomes and impacts (Figure 2).

Furthermore, critical components for successful implementation include customizing data to provide generative AI with operation-specific information, ensuring comprehensive user education, establishing strong governance frameworks, and developing robust methods for evaluating the impact post-implementation. This includes measuring ROI and closely monitoring usage patterns to ensure sustained performance and value realization.

Figure 2. Key points when implementing generative AI

The following are four critical considerations for implementing generative AI in the financial sector within the Southeast Asian context:

1) Security
The financial industry deals with highly sensitive information, such as customer financial data, necessitating a strong focus on security and data protection. Financial institutions must establish robust security measures to prevent data leaks and develop clear governance structures that limit the type of information users can input during operations. ABeam Consulting has extensive experience in building secure environments across multiple cloud platforms, and we provide support in formulating user governance structures and operational rules to ensure compliance and safety.

2) Addressing AI "Hallucinations"
One significant risk when deploying generative AI is the potential for "hallucinations," where the AI generates inaccurate or misleading information. In operations requiring high levels of precision, it is essential to mitigate these risks. Companies should implement frameworks where human oversight can detect errors or use technologies such as Retrieval-Augmented Generation (RAG) to ensure AI responses are backed by external, reliable data sources. For example, in a project aimed at streamlining customer support for an insurance company, we incorporated a review process where human agents verify the accuracy of AI-generated email content (see 4(3) ABeam LLM Partner implementation case study).

3) Compliance with Local Regulations
Different countries impose varying legal restrictions that can affect AI deployment. For instance, in Indonesia, regulations prohibit storing personal data overseas, limiting the range of applicable AI models. At ABeam Consulting, we assess the best generative AI models, including open-source options, to ensure compliance with local laws while meeting operational requirements. This approach enables us to select models that align with both regulatory guidelines and business needs.

4) Handling Multiple Languages
In Southeast Asia, the ability to support multiple languages is often a critical system requirement. Ensuring the accuracy and relevance of AI responses across languages requires careful evaluation and continuous improvement. Companies must decide on the optimal language structure for their knowledge base*, whether it involves unifying input and output in a single language or creating a separate knowledge base for each language. In some cases, it may even be necessary to use specialized generative AI models tailored to specific languages. ABeam Consulting provides solutions to help clients navigate these complexities, ensuring their AI systems are suitable for multilingual operations.

*Knowledge Base: A database containing company-specific reference information to enable generative AI models to provide accurate responses.

Generative AI offers immense potential for automating operations that were once considered too complex for machine-based solutions. However, successful implementation requires a combination of deep operational knowledge, AI expertise, and technological infrastructure. Many companies that have attempted to adopt generative AI without the right strategic approach have not seen the expected efficiency gains. ABeam Consulting combines its extensive financial industry insights with technical proficiency in AI and cloud technologies to deliver tailored solutions that address core business challenges.

3. ABeam Consulting’s Generative AI Solutions

ABeam Consulting offers a comprehensive range of services tailored to meet the specific challenges faced by clients at various stages of generative AI adoption. Below is a selection of our key offerings:

1) Business Value and Customer Experience Enhancement Program
Our “Business Value/Customer Experience Enhancement Program” includes study groups and workshops designed to help clients understand and leverage generative AI in their operations (Figure 3). Many organizations know they want to integrate AI into their business processes but struggle to identify where and how it can add value. This program helps clients pinpoint high-impact use cases by combining knowledge of generative AI capabilities with insights into existing operations.

The program starts with workshops grounded in design thinking, where we introduce participants to the latest trends, capabilities, and risks associated with generative AI. From there, we collaborate with clients to develop prototype applications and evaluate the feasibility of the generated ideas. This process ensures that the solutions we propose directly address business challenges rather than simply seeking to "implement AI for the sake of it."

Additionally, system environments and data management are critical factors in ensuring successful AI integration. Following the workshops, we offer further support in identifying system and data challenges, formulating IT strategies, and creating roadmaps to guide AI implementation efforts based on the ideas generated during the sessions.

Figure 3. Visualization of how a Business Value/Customer Experience Production Support Program would be held

2) ABeam LLM Partners

A common challenge with generative AI is its tendency to generate generalized answers derived from its training data, which may not be tailored to the specific needs of the business. To overcome this, ABeam LLM Partners provides a generative AI solution that customizes responses based on the client’s unique documents and data (Figure 4). This approach enables the application of generative AI to highly specialized and business-critical tasks, ensuring that the AI’s outputs are not only accurate but also aligned with the specific operational requirements of the client.

Figure 4. ABeam LLM Partners

For example, by centralizing past internal reports, generative AI can reference them as a knowledge base, enabling it to draft and review documents and answer user queries based on historical data. This reduces labor hours, enhances quality, and minimizes dependency on individual expertise.

When implementing ABeam LLM Partners, we recommend starting with a Proof of Concept (PoC) to assess the system’s potential impact. We can deliver a PoC in as little as 1.5 months. By utilizing the trial application during this phase, companies can evaluate the solution’s qualitative and quantitative benefits in real-world operations (Figure 5).

Figure 5. Example schedule for the introduction of ABeam LLM Partners

ABeam LLM Partner Implementation Case Study
Insurance Company: Automating Email Generation to Streamline Customer Support Operations

Figure 6. Visualization of a generative AI application

Overview

This section presents case studies where ABeam Consulting has implemented generative AI solutions to streamline customer support operations for insurance companies in Southeast Asia (Figure 6). We successfully optimized responses to complex customer inquiries—previously challenging to handle with AI or ChatGPT—by building systems that generate tailored replies based on customer-specific details, such as contract and product information.

Key Points and Anticipated Impact of the Initiative

In this project, ABeam Consulting deployed generative AI as an assistant within the customer support department. Given the current limitations of generative AI in producing 100% accurate responses, we designed a process where customer service staff review and refine AI-generated drafts before sending them.

To further enhance accuracy, we implemented a system that displays the documents and text references used by the AI when drafting emails, allowing staff to evaluate the appropriateness of the AI’s responses. This hybrid approach ensures maximum operational efficiency without compromising the quality of customer interactions, even with the potential for AI to generate incorrect outputs.

Additionally, we incorporated a feedback feature, enabling users to provide input on AI-generated responses (Figure 7). Continuous quality improvement through trial and error is essential for refining generative AI solutions, which is why ABeam supports the full process—from feedback aggregation to monitoring and enhancing response accuracy.

As a result of these efforts, this initiative achieved approximately an 80% reduction in customer email response workload after multiple rounds of quality improvement.

Figure 7. Feedback screen

ABeam Consulting provides comprehensive, end-to-end support for financial institutions, from identifying key areas where generative AI can be effectively applied to executing these projects. Leveraging our extensive experience in implementing generative AI solutions across Asia, including Japan, we help clients achieve meaningful operational improvements. We welcome companies interested in exploring the potential of generative AI to reach out for discussions and insights.

Reference: SEA-LION - AI Singapore

Insights TOP

Contact

Click here for inquiries and consultations