LLM Power Struggle: Who Will Reign in Large Enterprises?

Generative AI (GenAI), especially LLMs, have transitioned from buzzwords to essential business assets. Leading organisations have moved beyond the proof of concept (POC) phase and are now actively deploying GenAI solutions. Many non-technology companies have realised that developing proprietary LLMs is unnecessary for solving internal challenges . Instead, they acquire LLMs from vendors or the open-source marketplace.

This shift has ignited a crucial debate: who should own these new assets? Are LLMs data assets or IT assets? This article will explore this issue and reveal how industry leaders navigate this challenge. It offers actionable insights and strategies to ensure these roles deliver measurable business value and drive organisational success.

The ownership of the acquired LLMs

Though there is no set principle regarding this, when an organisation procures an LLM, ownership should ideally fall under the CIO or CTO rather than the Chief Data and Analytics Officer (CDAO).

The CTO focuses on developing and implementing technology solutions, including R&D and product development. The CTO should oversee the development of new products and services or the enhancement of existing technological capabilities using LLMs. The CTO’s role in driving technological innovation aligns well with the strategic use of LLMs. At Tesla, the CTO oversees the development of AI technologies integral to the company’s products, such as autonomous driving systems. This highlights the importance of the CTO managing advanced technological assets like LLMs.

On the contrary, the CIO is responsible for IT infrastructure, including software systems, cybersecurity, and technology strategy. Externally procured LLMs need to be integrated into the existing IT ecosystem. This involves ensuring compatibility with current systems, managing software licenses, and maintaining security protocols. The CIO’s expertise in managing these aspects makes them the ideal owner. At JPMorgan Chase, the Global CIO, Lori Beer, oversees the integration and deployment of AI technologies such as the LLM Suite. This ensures that the LLM is effectively utilised across various departments to enhance operational efficiencies and drive business value.

The CDAO is primarily focused on data strategy, governance, and analytics. Their responsibility is to leverage data to drive business insights and decisions. While the CDAO plays a crucial role in utilising the outputs of LLMs for data analysis and insights, they may not have the necessary expertise in managing the IT infrastructure and integration processes required for externally procured LLMs. This can lead to inefficiencies and potential security risks. For example, at JPMorgan Chase, while the CDAO is involved in using their LLM Suite to enhance data analytics, the actual ownership and integration are managed by the CIO. The appropriate role handles the technical and infrastructural aspects, ensuring proper separation.

Is the acquired LLM an IT asset or a data asset?
When an organisation acquires LLMs from vendors or the open-source marketplace, it can be argued for being both an IT asset and a data asset, depending on its use and management.

Arguments for IT asset
1. Integration and Infrastructure: The LLM is a sophisticated piece of software that requires robust IT infrastructure for deployment, maintenance, and integration with other systems. This includes hardware resources, software licenses, and IT management.
2. Security and Compliance: Managing an LLM involves ensuring compliance with security protocols and regulatory requirements. This is typically within the domain of the IT department, which is responsible for safeguarding the organisation’s technological assets.

Arguments for data asset
1. Data Utilisation and Insights: The LLM is trained on vast amounts of data and generates valuable insights, making it a critical data resource. It can be applied to analyse, interpret, and generate data, contributing significantly to the organisation’s data strategy.
2. Value Generation: LLM’s primary value lies in its ability to process and generate data, which can be adapted to improve business processes, customer interactions, and strategic planning.

Let’s draw inspiration from JPMorgan Chase, where the Global CIO, Lori Beer, manages the integration of LLM Suite into the company’s IT ecosystem. This strategic oversight ensures the LLM is effectively used across various departments, enhancing operational efficiencies, and driving business value. Under Lori Beer’s leadership, the LLM Suite adheres to the bank’s stringent security standards, protecting sensitive data and maintaining compliance with financial regulations. While the CIO manages the technical aspects, the data and analytics teams leverage the insights generated by the LLM Suite to enhance decision-making and drive business insights. The LLM Suite is expected to deliver significant value, potentially up to $2 billion, particularly in areas like fraud prevention. This underscores its role as a data asset that generates substantial business benefits.

Classifying an LLM as an IT or a data asset depends on the organisation’s structure and strategic goals. This dual classification underscores the importance of a collaborative approach, where IT and data teams work together to maximise the benefits of LLMs.

How LLMs are different from any other IT asset, say CRM software?
CRM and LLMs leverage data to drive business value and support strategic decision-making. CRM software uses structured customer data to provide insights into customer behaviour, sales trends, and marketing effectiveness. LLMs can analyse vast amounts of unstructured data to understand language patterns and generate meaningful responses. Both tools enhance decision-making, automate tasks, and personalise interactions, making them valuable assets for improving efficiency and driving business success.

However, CRM software is typically not considered a data asset because it primarily serves as a tool for managing and utilising customer data rather than a data source. In contrast, LLMs can be seen as data assets because they generate new insights and knowledge from the data they process, adding intrinsic value to the organisation.

Concluding remark
The question of ownership of LLMs reflects a broader shift in how enterprises perceive and manage their technological assets. Whether seen as IT tools driving integration and infrastructure or data resources unlocking insights, the value of LLMs lies in their capacity to bridge the gap between innovation and business impact. This duality necessitates a collaborative approach among IT, data, and analytics teams to maximise their potential. As exemplified by organisations like JPMorgan Chase, clearly defined roles—whether under the CIO or the CTO—enhance the strategic utilisation of LLMs, ensuring security, compliance, and measurable business outcomes. The CDAO, while crucial in leveraging the data and insights generated by LLMs, should not be responsible for their ownership due to the technical and infrastructural demands involved. In an increasingly AI-driven world, enterprises must view LLMs as catalysts for transformative growth, not merely as assets.

References:

https://www.linkedin.com/pulse/large-language-models-business-how-make-right-dr-m-maruf-hossain-phd-pqpge/