Leveraging Generative AI to Accelerate Finance Transformation in Banking.

Since the launch of advanced generative AI technologies like ChatGPT in late 2022, businesses across all industries, especially banking, have found themselves at the forefront of a technological revolution. Every boardroom is asking about the risks and opportunities that Generative AI (Gen AI) presents to their business. Gen AI will reshape how companies operate, from internal efficiencies to customer engagement, and holds immense potential for the banking sector, offering the ability to automate and enhance many core functions, streamline operations, and unlock new business value.

Disclosure: This blog was created based on a conversation (or series of prompts) with Revvy, a ChatGPT developed by Revvence and trained on finance transformation and systems change-related content.

Generative AI: The New Frontier in Banking

In the banking industry, up to 40% of all working hours involve tasks that can be significantly impacted by large language models (LLMs) like GPT-4. 

Routine language tasks comprise approximately 62% of employee work hours and present a prime opportunity for productivity improvements. Through AI-driven automation and augmentation, banks can reduce the time spent on repetitive tasks and refocus resources on higher-value activities like strategic decision-making and innovation.

The banking sector faces unique challenges, including stringent regulatory requirements, rising operational costs, and the growing need to improve customer service while maintaining data privacy. Generative AI is poised to address these challenges by transforming processes like customer service automation, risk management, RWA optimisation, and regulatory compliance.

While 98% of banking executives believe that generative AI will be crucial to their strategic roadmap in the next 3 to 5 years, many institutions are still cautiously experimenting with this technology to evaluate potential risks and integration challenges, but they can’t afford to move too slowly.

Why AI Transformation Will Outpace Digital Transformation in Banking

The speed at which AI transformation occurs in banking far exceeds the pace of previous digital transformations. While digital transformation was revolutionary, banks needed years to modernise legacy systems, integrate cloud solutions, and adopt digital-first strategies. In contrast, AI transformation, particularly with generative AI, is happening much faster, driven by several key factors:

Pre-existing Digital Infrastructure: 

One of the most significant enablers of AI transformation is that many banks have already undergone extensive digital transformations. With cloud infrastructure, digital data flows, and automated processes in place, banks can now build on these foundations to implement AI capabilities without overhauling legacy systems first.

 Exponential Growth in AI Capabilities: 

AI technologies, particularly generative AI, have rapidly advanced in just a few short years. Unlike digital transformation, which requires slow, stepwise improvements in infrastructure, AI advancements can be deployed quickly across existing systems. Oracle’s AI tools, for instance, can be integrated directly into existing applications, accelerating the adoption of AI-driven processes without extensive reconfiguration.

Data Readiness: 

AI thrives on data, and today’s banks have vast amounts of structured and unstructured data collected over years of digital transformation efforts. This wealth of data can now be leveraged to train AI models and generate insights that were not possible during the earlier stages of digital transformation. By refining their data strategies, banks can more rapidly implement AI solutions to enhance decision-making, improve risk management, and optimise customer interactions.

Customer and Regulatory Pressure: 

External forces also drive AI transformation in banking. Customers expect more personalised, faster, and seamless banking experiences, which AI can provide through intelligent automation and enhanced customer support tools. At the same time, regulatory bodies are increasingly encouraging the use of AI for compliance, fraud detection, and risk management, giving banks further incentives to accelerate AI adoption.

AI as an Industry Differentiator: 

Unlike digital transformation, which was necessary for banks to remain competitive, AI is quickly becoming a differentiator. Early adopters of AI, particularly in banking, are expected to outperform their competitors by improving operational efficiency, reducing costs, and offering innovative products and services. Banks that quickly adapt to AI capabilities will gain significant competitive advantages in areas like automated financial advising, enhanced customer experiences, and predictive analytics for investment and lending decisions.

As AI continues to evolve rapidly, banks that have already laid the groundwork during the digital transformation era are uniquely positioned to benefit from the speed of AI transformation. With Oracle’s powerful AI tools, financial institutions can quickly scale AI across their operations, revolutionising everything from customer service to regulatory compliance, optimising RWAs, improving product margins and automating vast amounts of reporting to position themselves as leaders in the future of banking.

Preparing for Generative AI in Banking

Banks must take several foundational steps to harness Gen AI's power fully. Here's how financial institutions can prepare for the adoption of generative AI technology:

1. Develop a Comprehensive AI Strategy: 

Banks need a well-defined strategy to identify which AI platforms and LLMs align with their business needs. For example, institutions may consider using off-the-shelf pre-trained models, customising them for internal operations, or using both approaches. This topic deserves a blog of its own, but as a starter, we've created a Metrics-Driven Approach to Generating AI Use Cases. 

2. Create Secure AI Workflows: 

Implementing secure, compliant AI solutions is paramount in the banking industry. Institutions should establish trusted "sandboxes" for experimentation, where Gen AI can be safely tested and refined to ensure regulatory compliance and customer data security.

3. Optimise Data for AI: 

High-quality, standardised data is essential for generative AI to deliver accurate insights. Banks need to ensure their enterprise data is well-structured and up to date, as data inconsistencies can lead to unreliable AI outputs, particularly in areas such as credit risk analysis and anti-money laundering (AML) processes.

Banks must also realise that striving for the perfect data set isn’t feasible. Trying Gen AI use cases will accelerate the creation of high-quality data sets.

How Gen AI will Transform Banking Operations

Generative AI’s potential for transformation in banking spans many areas. Our blog has deep-dived into several specific areas. You can read some of the blog content specific to AI here.

At a high level, here are some critical banking use cases where generative AI is already making an impact:

Risk Management and Compliance: 

AI tools can analyse large amounts of structured and unstructured data to enhance risk management processes, including real-time fraud detection, regulatory compliance checks, and transaction monitoring.

Customer Experience: 

Integrating generative AI into customer-facing operations allows for the automation of routine tasks, such as responding to inquiries, processing transactions, and personalising customer interactions through AI-driven insights. This frees up human agents to focus on complex customer needs.

Operational Efficiency:

Automating back-office tasks such as document processing, account reconciliation, and financial reporting can significantly reduce human errors and increase the speed of operations, helping banks improve service delivery while cutting costs.

Data Analytics and Decision-Making:

Generative AI can quickly sift through massive amounts of data to generate actionable insights, assisting banking executives in making informed decisions on lending policies, investment strategies, and portfolio management.

The Gen AI Platform Choices for Banks​​

In the banking industry, it is common for institutions to operate multiple platforms from vendors like Oracle, SAP, Workday, AWS, Azure, and GCP, each serving critical functions within their financial ecosystems. As these platforms continue to evolve, all of these major vendors are integrating AI capabilities directly into their offerings. This approach means that banks don't need to adopt an entirely new AI system; instead, they should focus on leveraging the AI features embedded within their existing infrastructure.

In this context, AI is not a standalone product but a feature that enhances the platforms already in place. For example, Oracle embeds AI into its cloud services to improve operational efficiencies, while AWS, Azure, and GCP offer AI tools as part of their cloud computing services, specifically designed to work within existing enterprise frameworks. The key for banks is to harness these built-in AI capabilities across their multi-vendor environments —integrating AI where it makes the most sense operationally without requiring a significant overhaul of their systems. By doing so, banks can adopt AI incrementally and strategically, improving their existing processes while staying agile and scalable.

The remainder of this blog focuses on Oracle’s AI capabilities for the banking industry.

Oracle's Gen AI Capabilities for Banking

Oracle’s AI infrastructure, built on Oracle Cloud Infrastructure (OCI), is ideally suited for financial institutions looking to adopt generative AI at scale. OCI provides one of AI's most advanced cloud infrastructures with its low-cost, high-performance GPU clusters. This enables banks to process and analyse large datasets faster and more cost-effectively than alternative enterprise offerings.

Some of the key differentiating features we think are important to consider include:

Trusted, Proven Models and Technology Partners 

Oracle provides banking institutions access to trusted large language models (LLMs) and AI tools optimised for mission-critical systems. Oracle has a partnership with Cohere, a leading generative AI company for enterprise-grade large language models (LLMs), and a strategy that delivers several significant benefits for banks:

> Choice of models for enterprise use cases

Banks can access pre-trained, foundational models from Oracle’s AI partners (such as Cohere R and Meta Llama 3) to summarise, embed, and generate text.

> Flexible fine-tuning

Banks can create custom models by fine-tuning base models with their data set. They can build new custom models or new versions of existing ones.

> Content moderation and controls 

Banks can control the usage and governance of AI models. This allows them to ensure that AI-generated content complies with regulations while also providing flexibility to adapt to new requirements or conditions.

> Dedicated AI clusters 

With dedicated AI clusters, banks can host foundational models on dedicated GPUs that are private to you. These clusters provide stable, high-throughput performance required for production use cases and can support hosting and fine-tuning workloads. OCI Generative AI also enables banks to scale their cluster with zero downtime to handle changes in volume.

The Depth and Breadth of the Offering.

Oracle’s AI strategy focuses on integrating AI capabilities across its cloud infrastructure, databases, and applications, providing powerful AI capabilities suitable to multiple banking use cases.

> AI Inside Applications

Popular applications used widely in banking, such as Oracle Fusion ERP and Oracle EPM Cloud, can surface AI-powered insights directly inside the apps with embedded, fully functional AI capabilities.

> Generative AI

Banks can choose from managed open-source or proprietary LLMs. They can fine-tune prebuilt models and augment them with their enterprise data, and they can take advantage of built-in vector databases (a critical feature for building internal Generative AI applications).

> AI services

Access a broad range of AI Services, such as Document Understanding and AI Agents and add prebuilt models to existing applications to leverage AI that can be customised with bank data for improved model quality.

> Generative AI and Machine Learning for data platforms

Collaboratively build, train, deploy, and manage machine learning (ML) models using existing open-source frameworks and data platforms, such as AWS, and leverage Oracle’s pre-built capabilities designed specifically for use cases such as stress testing, climate modelling and risk analysis.

> AI infrastructure

Oracle’s AI infrastructure is unmatched in the enterprise space. Its strategic relationship with NVIDIA gives it access to the highest number of the most powerful GPUs (the chips that power AI). Banks need high-performance AI, and Oracle’s Supercluster can scale up to 32,768 GPUs today and 131,072 GPUs soon.

Probably the most crucial feature of Oracle's AI strategy for banking is Soverign AI, which enables banks to deploy AI infrastructure in their own data centres or dedicated data centres operated by Oracle to help meet performance, security, and AI sovereignty requirements.

Oracle's Retrieval Augmented Generation (RAG) Capabilities

One of Oracle’s key innovations in the AI space is making Retrieval Augmented Generation (RAG) capabilities easier to deploy across enterprise workflows. RAG enhances LLMs by combining the power of generative AI with real-time data retrieval from enterprise data sources.  Here’s how RAG transforms banking operations.

Oracle’s RAG capabilities are integrated within Oracle Cloud Infrastructure (OCI) and supported by Oracle’s AI Services, which enable seamless interaction between generative AI models and your internal databases, especially the Oracle database. This solution is highly secure, scalable, and built to manage the complex, data-intensive tasks typically encountered in banking operations.

Workflow Example: CSRD Disclosure Compliance

To better understand the power of Oracle’s RAG solution, let’s walk through a real-world use case: complying with CSRD (Corporate Sustainability Reporting Directive) disclosure requirements.

1. Data Collection 

In this scenario, let’s assume that a bank uses AWS S3 as its core enterprise data platform. Oracle has a feature called Select AI in their Autonomous Database, which enables banks to easily add data for a CSRD disclosure into a RAG application without moving data from existing platforms like AWS S3. The Oracle platform can query data directly from S3 buckets, retrieving ESG data seamlessly. Banks can maintain their AWS infrastructure while leveraging Oracle's AI-driven capabilities to automate data collection and reporting. This integration simplifies compliance processes and ensures up-to-date data is used without complex migration.

2. Fine-Tune for CSRD 

Oracle’s Generative AI approach allows the bank to fine-tune models to handle specific compliance needs, such as CSRD disclosures. To fine-tune the RAG application the bank will first define the particular structure of the CSRD report by creating a template with sections for emissions, energy usage, and other sustainability metrics. This template acts as a framework for the AI model. Then, using the Oracle AI platform, the bank can fine-tune the LLM by training it with historical data and compliance documents relevant to CSRD, ensuring the model understands the content and structure. The LLM is taught to populate the template based on real-time data, ensuring accuracy and compliance with CSRD requirements. Since OCI’s GPU clusters power the models, they can process massive datasets quickly and efficiently.

3. Custom Report Generation  

After gathering the necessary data, Oracle’s RAG application automates the creation of CSRD disclosures by utilising the pre-trained LLM that understands the structure of the required template. As explained above, the LLM is trained with specific CSRD requirements and placeholders for metrics like carbon emissions, energy consumption, and sustainability targets. When a user prompts the system (e.g., "Generate CSRD report"), the LLM retrieves the relevant data and populates the template with real-time information, ensuring the final report is accurate, structured, and ready for submission. This process simplifies complex compliance reporting through automation.

Benefits of a RAG Application for CSRD Reporting

1. Time Efficiency: With Oracle’s RAG solution on OCI, your team no longer needs to spend hours pulling together data for CSRD disclosures. The AI retrieves and compiles everything in real time, significantly reducing the reporting cycle.
2. Accuracy and Consistency: By using the Oracle platform for data management, RAG ensures that the generated reports always draw from accurate, up-to-date information, reducing the risk of errors.
3. Adaptability: Should CSRD regulations be updated, Oracle AI Services automatically adjusts the report outputs. The streamlined process allows your bank to stay compliant without significant manual adjustments.

Conclusion

Oracle's Gen AI solutions are well-positioned to address the complexities of banking. With capabilities ranging from AI-embedded applications to advanced RAG-based solutions, Oracle can accelerate and reduce the cost of AI applications for banks in every step of their AI journey while ensuring compliance, data security, and scalability.

At Revvence, GenAI will play a large part in the technology and finance transformation work we deliver for our clients. Revvence has developed a unique GenAI application called Revvy. Based on OpenAI's ChatGPT 4o model, it is trained in finance transformation processes, best practices, and technology architectures.