The banker's guide to generative AI

Banks are racing to implement generative AI (GenAI), and the right operational framework can unlock its full potential. This cutting-edge technology transforms the banking sector by enhancing customer service chatbots, detecting fraud, and accelerating tasks like coding, preparing pitch book drafts, and summarizing regulatory reports.

The McKinsey Global Institute (MGI) projects that GenAI could contribute an annual value of $200 billion to $340 billion to the global banking industry, which translates to 2.8 to 4.7 percent of total sector revenues, primarily through improved productivity. However, as financial institutions rush to adopt this technology, they encounter several challenges. Properly deploying GenAI can unlock significant benefits, but mishandling it can result in complications.

Banks must develop strong capabilities across several key areas to achieve sustained value beyond initial proofs of concept. These areas include having a strategic roadmap, attracting and retaining the right talent, establishing an effective operating model, leveraging advanced technology, managing and utilizing data efficiently, and implementing robust adoption and change management practices.

Each of these elements is interdependent and requires cohesive alignment throughout the organization. For example, even the most well-designed operating model will fail to deliver results if it lacks the necessary talent or access to quality data. This holistic approach ensures that all components work together to maximize the potential of generative AI in the banking sector.

Benefits of the operating model

A centrally managed generative artificial intelligence operating model offers multiple advantages.

• Firstly, given the limited availability of top-tier GenAI talent, centralization enables the organization to deploy its talent more effectively, ensuring that the entire enterprise benefits. This approach also allows for the creation of a cohesive, world-class GenAI team that fosters camaraderie, thereby attracting and retaining top talent.
• Secondly, in a dynamic environment where new large language models and GenAI features are constantly emerging, a centralized team can more effectively monitor and adapt to these changes compared to multiple dispersed teams.
• Lastly, a centrally managed model is particularly valuable in the early stages of an organization's GenAI initiative. During this phase, it is crucial to make frequent and significant decisions regarding funding, technological architecture, cloud providers, large language model providers, and partnerships. A central team can streamline these decisions, ensuring consistency and strategic alignment.

Why operating model is key

An operating model delineates how a company functions, encompassing its structure (including roles, responsibilities, governance, and decision-making), processes (performance management, systems, and technology), and people (skills, culture, and informal networks). Financial institutions that have successfully harnessed generative AI have developed tailored operating models specifically designed to address the unique characteristics associated with this new technology rather than trying to fit GenAI into their pre-existing frameworks.

It has been observed that most financial institutions effectively leveraging GenAI are adopting a centrally led operating model for this technology, even if other parts of the organization remain decentralized. This approach is likely to evolve as GenAI technology matures.

The optimal operating model for a financial services company's GenAI initiative should facilitate scalability while aligning with the company's organizational structure and culture; there is no universal solution. An appropriately designed operating model, adaptable as the institution grows, is essential for effectively scaling GenAI initiatives.

In essence, a suitable operating model enables a financial institution to efficiently carry out three types of activities:

• Strategic guidance: Identify clusters or domains of generative AI use cases that align with the enterprise’s strategic objectives. Prioritize these use cases into a roadmap that maximizes value while managing risk and monitoring value creation to ensure efficient resource allocation.
• Standardization: Define common standards, such as those concerning technology architecture choices, data practices, and risk frameworks and controls. These standards increase efficiency and allow insights from completed projects to be applied to new ones.
• Implementation: Design and test the technical solutions for use cases, implement the use cases that meet performance and safety criteria, and scale them if there is a valid business case. Ensure that their impact is tracked and delivered effectively.

Structuring GenAI operating models in banking

Banks and other financial institutions can adopt various approaches to structuring their generative AI operating models, ranging from highly centralized to highly decentralized frameworks.

A recent McKinsey review of GenAI usage among 16 of the largest financial institutions across Europe and the United States, representing nearly $26 trillion in assets, revealed that over 50 percent have opted for a more centrally led organization for GenAI.

This is notable even in instances where their typical setup for data and analytics is relatively decentralized. This centralization is expected to be a temporary measure, with the structure likely becoming more decentralized as the technology matures. Eventually, businesses might find it advantageous to allow individual functions to prioritize GenAI activities according to their specific needs.

Essential considerations for implementing GenAI operating models

Implementing a GenAI operating model requires financial institution leaders to make crucial decisions across various areas.

This checklist can help executives devise the optimal operating model for their organizations:

• Strategic direction: Identify the leaders responsible for defining the GenAI strategy and deciding whether it will be established enterprise-wide or at the business unit level. This includes envisioning the potential value and assessing which functions or processes will be most impacted by GenAI.
• Domain and use case identification: Determine who will define the enterprise domains or clusters of GenAI use cases and identify specific use cases within those domains.
• Deployment strategy: Decide on the deployment model for implementing domains and use cases, choosing between being a "taker" (procuring targeted solutions from vendors), a "shaper" (integrating broader solutions from vendors), or a "maker" (developing in-house solutions to reshape core business functions).
• Funding structure: Establish how GenAI use cases will be funded, considering the level of centralization or decentralization in the GenAI approach. This often involves a combination of funding from individual business units and a central GenAI team.
• Talent acquisition and development: Identify the necessary skills for GenAI initiatives and implement strategies for hiring, upskilling, strategic outsourcing, or a combination of these approaches. Additionally, define the role of "translators" who bridge the gap between business needs and technical requirements.
• Change management: Form a committee to lead the execution of a change management plan, ensuring the required shifts in mindsets and behaviors for successful GenAI adoption across the organization.

To select the best operating model, financial institutions must address key considerations, such as setting clear expectations for the GenAI team's role and embedding flexibility to allow the model to evolve over time. This flexibility should encompass both high-level organizational aspects and specific components like funding.

The dynamic nature of GenAI in banking necessitates a strategic approach to operating models. Financial institutions must balance speed and innovation, adapting their structures to fully leverage the technology.