With the advent of Generative AI, artificial intelligence has entered mainstream consumer awareness, sparking global recognition of the technology's immense potential. Banking and financial services firms are continuing to accelerate their GenAI adoption trajectories. NatWest and OpenAI have joined forces to enhance the lender's digital assistants and customer support processes, Temenos is bringing high-performance, on-premises generative AI, and at JPMorgan Chase, AI is reshaping business practices.
Clearly, banks see GenAI as a powerful tool to drive innovative products and services. They are delving into transformational use cases, including algorithmic portfolio optimization, automated customer support, personalized financial planning, and dynamic pricing models to revolutionize both operational efficiency and customer engagement.
At the same time, as with any much-hyped technology, GenAI comes with its own set of challenges: use case prioritization, including algorithmic portfolio optimization; compliance risks; model hallucination; privacy concerns; data copyright issues; biases; and misinformation. Business leaders are now thinking about the next wave of evolution in GenAI – what’s the best operating model, how to manage data, and how to deploy faster while managing risk?
Banking on GenAI: A Hybrid Operating Model
The most common operating model for banks aiming to implement GenAI involves a combination of a centralized AI Governance Council, a centralized GenAI team, and federated execution pods. The AI Governance Council is responsible for providing oversight, reviewing business cases, and sharing best practices. The centralized GenAI team provides common tools, data infrastructure, governance, data science, and data engineering capabilities while promoting best practices. Composed of business stakeholders, data scientists, and data engineers, the federated execution pods are designed to rapidly build and implement use cases in close collaboration with business units. This operating model ensures centralized access to governance, prioritization, and shared best practices while delegating decision-making and prioritization authority to individual business units; thus, it offers scale, speed, and a federated rapid execution model.
Building a Bank-centric GenAI Strategy
Over the years, banks and financial institutions have dedicated substantial investments to constructing algorithms and machine learning models for market differentiation. Today, with best-in-class GenAI models equally accessible to every enterprise, the true differentiator lies in how enterprise data is leveraged to train these models. The emphasis is shifting from building models to establishing a robust data foundation that uniquely and strategically trains large language models (LLMs).
Recently, the development of “sovereign AI” has emerged, referring to a nation's ability to produce artificial intelligence using its own infrastructure, data, workforce, and business networks. When applied to the financial services sector, this concept enables banks to utilize their own data and networks to harness GenAI capabilities while adhering to local regulations. Additionally, the integration of agentic AI can significantly enhance banks' operations by enabling autonomous decision-making, improving predictive analytics, and facilitating proactive customer engagement. Agentic AI models can independently analyze market trends, anticipate customer needs, and initiate personalized interactions, thereby driving efficiency and innovation in banking operations. A prime example of this is Wipro’s initiative to develop AI-powered digital assistants that operate within a client’s infrastructure, furthering the goals of sovereign AI.
The Imperative of Establishing a Robust Governance Framework
In the rapidly evolving banking landscape characterized by M&A activities and increasing regulatory scrutiny, data quality stands as a non-negotiable cornerstone for sustainable growth and compliance. As a heavily regulated sector, every facet of banking, from customer segmentation to compliance reporting, necessitates a rigorous data management and governance framework. The imperative for a robust data foundation becomes even more pronounced when integrating disparate data ecosystems post-M&A, requiring meticulous attention to data integrity, consistency, and traceability.
To help, banks are increasingly adopting the National Institute of Standards and Technology (NIST) AI Risk Management Framework to ensure responsible AI practices. This framework helps financial institutions establish accountability, transparency, and compliance with evolving regulatory requirements, while embedding ethical commitments and operational standards into AI development.
By aligning internal principles with external frameworks like NIST, banks can deploy AI tools that are trusted, explainable, and compliant, thereby safeguarding corporate reputation and accelerating innovation.
In addition to laying a robust data foundation, governance of AI models emerges as a critical imperative for banks aiming to harness the transformative potential of GenAI. The following considerations underscore the importance of a structured governance framework tailored for Generative AI applications:
- Information Security: To safeguard confidential information, it is paramount to restrict direct access to GenAI interfaces such as ChatGPT. Banks should instead deploy purpose-built business applications that leverage Generative AI models within a fortified information security framework.
- Purposeful Restriction: The ubiquitous nature of general-purpose GenAI models, trained on vast and diverse data sets, necessitates stringent measures to ensure that responses align with specific business objectives. Banks should develop specialized applications atop Generative AI models, programmatically evaluating each request's relevance to the intended purpose before generating a response
- Custom Training and Fine-tuning: To harness the full potential of GenAI, banks can leverage proprietary knowledge through custom training or fine-tuning of models. This approach not only enhances transparency and credibility but also facilitates the creation of a robust framework for traceability, enabling banks to trace responses back to the source documents used for training.
- Response Moderation: As banks integrate Generative AI into diverse use cases, from virtual assistants to marketing initiatives, implementing rigorous response moderation becomes indispensable. This involves deploying advanced content detection mechanisms and leveraging custom-trained classification models to identify and mitigate potentially harmful elements in generated responses, including plagiarism or copyrighted content, and bias.
Wipro’s Enterprise Generative AI Studio (WeGA) accelerator provides a tailored and vigilant approach for establishing guardrails to govern AI models. By adopting WeGA, banks can effectively navigate the complexities of Generative AI while ensuring alignment with regulatory mandates and driving sustainable innovation, thanks to a structured governance framework that encompasses information security, purposeful restriction, custom training, and response moderation.
The Outlook for Banks and Financial Institutions
With a clear focus on building a robust data foundation and designing an operating model that balances centralized capabilities with federated use case prioritization and execution, banks are setting a strong foundation. By also establishing proper governance mechanisms, they are well-positioned to effectively capitalize on the potential of the GenAI revolution.
Banks can envision a future in which GenAI models seamlessly analyze new regulatory documents in real time, automate updates to training materials, and conduct sentiment analysis on corporate customers to inform underwriting decisions. Furthermore, these models hold the promise of enhancing customer service interactions, facilitating rapid response times, and empowering representatives to address more complex queries. The evolution of virtual banking assistants, characterized by human-like responsiveness and "always on" availability, further underscores the transformative potential of GenAI.
However, it is crucial to recognize that achieving reliable GenAI capabilities transcends the user-friendly allure of platforms like ChatGPT. Building a secure, accurate, and responsible enterprise-grade GenAI function demands meticulous planning, customization, and alignment with industry-specific data and regulatory frameworks. While the journey towards realizing the full potential of GenAI in banking may be complex and nuanced, the rewards — enhanced efficiency, customer experience, and innovation — are undeniably worth the investment and strategic focus.
