Over the past year, much of the initial enthusiasm surrounding generative AI (gen AI) has given way to more of a sentiment of cautious optimism. In particular, organizations seeking to recalibrate their AI strategies have come to realize that effectively leveraging gen AI demands a type of effort that is characterized by more than just enthusiasm— it requires a solid foundation of information governance best practices and a sustained commitment to managing information and records that is both compliant and efficient.
We believe that by focusing on effective information governance, organizations can ease their transition from being mere users of generative AI to creators and managers of accurate and effective large language models (LLMs).
Capturing the potential value of generative AI hinges on robust information governance. As companies strive to harness the power of gen AI, they must remember the hard lessons learned from previous digital and AI transformations. Building competitive advantage requires developing organizational and technological capabilities that enable broad innovation, deployment, and continuous improvement of solutions at scale. This involves rewiring business processes to accommodate distributed digital and AI innovation effectively.
A clear and broadly formulated strategic information governance vision is crucial for any organization aiming to advance its gen AI capabilities and to improving its AI LLMs. An important information governance tool in this respect is the creation of cross-functional product teams similar to a typical information governance steering committee (which is composed of diverse business units and functional areas) that shared objectives and incentives to ensure that gen AI tools are built and implemented effectively. These teams should include members from various departments including IT, Compliance, Legal, HR, Facilities, Finance, and other core functional areas, ensuring diverse perspectives and expertise are leveraged.
Another important information governance strategy is to use metadata management principles and tools to ensure that data is accurately labeled, making it easier to find, retrieve, and use. This is especially important for training LLMs, which require large volumes of high-quality, well-organized data. A study by Gartner, for example, found that organizations that effectively manage their metadata see a 15% improvement in data utilization and a 10% reduction in operational costs. Effectively managing metadata helps ensure that data used to train AI models is easily accessible and reliable, which is essential for developing accurate and effective LLMs.
Retention and privacy compliance are also critical components for developing effective LLMs. Information governance principles help to ensure that records and data are retained for the appropriate amount of time and that privacy regulations are followed to help mitigate risks associated with data breaches and non-compliance. According to a recent report by IBM, the average cost of a data breach in 2023 was $4.45 million. By implementing strong retention and privacy compliance policies, organizations can protect themselves from such costly incidents. Moreover, compliance with privacy regulations is essential when handling the vast amounts of data required for training LLMs, ensuring that personal information is not misused or mishandled.
Upskilling the workforce is another essential aspect of transitioning from users to creators of LLMs. Establishing a data and AI academy can help upskill employees across the organization. Training dispatchers, service operators, and other relevant staff ensures they can effectively work with data and gen AI tools. A survey by PwC found that 74% of CEOs are concerned about the availability of key skills, emphasizing the need for continuous employee training and engagement. By investing in employee training, organizations can build a workforce that is proficient in using and developing gen AI tools, thus driving innovation and productivity.
High-quality content is essential for customizing LLMs. This is where information governance concepts of version control come in. Companies need a systematic method for capturing and managing digital content to ensure it is accurate, timely, and not duplicated – and that LLMs are “fed” the right, final versions of data! An important aspect of managing generative AI content is ensuring quality. Generative AI is widely known to "hallucinate" on occasion, confidently stating facts that are incorrect or nonexistent. Errors of this type can be problematic for businesses but could be deadly in healthcare applications. Ensuring content accuracy through rigorous quality assurance processes is crucial.
By focusing on strategic vision, cross-functional teamwork, upskilling, selecting appropriate technology, and robust data management, organizations can move from being effective users of generative AI to creators of accurate LLMs. As AI technology evolves rapidly, companies must be prepared to adapt their approaches continuously, ensuring they remain at the forefront of innovation and competitive advantage. The journey from user to maker of LLMs is challenging but offers substantial rewards in enhanced productivity, innovation, and strategic growth.
Comments