Many organizations face significant barriers in deploying generative AI primarily due to unprepared data. Only 18% of companies report that their data is fully ready for AI deployment. Key issues include the need for data organization, tagging, and governance to support AI integration. Ongoing communication between data engineers and subject matter experts plays a crucial role in maintaining data quality and addressing issues after AI deployment. Experts recommend a proactive approach to data governance and ethical considerations for data exposure when pursuing AI readiness.
The deployment of generative artificial intelligence (AI) poses significant challenges for many organizations, primarily due to the lack of adequately prepared data. While there is enthusiasm from boards of directors towards adopting AI technologies, Chief Information Officers recognize that the situation is more complex than presenting a viable AI use case. Prukalpa Sankar, co-founder of Atlan, underscores that successful AI implementation hinges on the readiness of data, stating, “Everybody’s ready for AI except your data.” A recent survey involving over 1,300 technology and data executives revealed that only 18% of organizations consider their data fully ready for AI deployment, with another 40% indicating they are mostly prepared. To transition into a state of readiness, companies must tackle various challenges, including the consolidation and organization of disparate data sources, a task primarily managed by data engineers. Sankar emphasizes the necessity of integrating data previously isolated within different business units to facilitate specific AI use cases. Moreover, intricate processes of data tagging and classification must be completed to ensure compliance with privacy regulations. As Sankar illustrates, “Depending on who’s asking the question, I can change the data that goes behind it.” The governance of data becomes increasingly complex with AI integration, as Matt Carroll, CEO of Immuta, identifies that traditional governance models cannot adequately address the dynamic needs introduced by AI technologies. Carroll notes, “When you think of traditional business intelligence… governance was always a structured, well-oiled machine.” The implementation of an effective data governance framework is crucial for organizations embarking on AI initiatives. Carroll identifies three essential elements for AI readiness: accessibility to data, utilization of that data, and the capacity to monitor its usage. Currently, sectors such as government and finance are focusing on data governance, trust, and security more intensely than others, but Carroll advocates for broader adoption of these practices across all industries handling sensitive information. To maintain ongoing data readiness post-deployment, Carroll suggests implementing feedback loops between domain experts and data engineers, such as an AI hotline. This facilitates swift identification and rectification of issues such as erroneous data tagging or model inaccuracies. Initiatives like continuous model testing will ensure adherence to company quality standards, enhancing the overall performance of deployed AI systems. As companies begin their AI deployment journey, they are also devising metrics to assess AI readiness—such as AI readiness scores—aggregating relevant performance indicators. Another emerging trend involves assigning dual roles where employees become responsible for data stewardship alongside their primary job functions. Carroll emphasizes the increasing need for specialized data governance roles within organizations. Sankar aptly compares the data infrastructure landscape to a marketplace, with AI use cases on one side and the complex data structure on the other. For organizations serious about AI implementation, preparing data is paramount. However, it is critical to approach this with ethical considerations, as Carroll encourages decision-makers to assess the implications of exposing certain data types in their systems prior to advancing toward AI readiness. Only after these decisions can companies genuinely pursue AI deployment.
The integration of generative AI technologies into business practices is increasingly being prioritized by company leadership. However, a significant barrier to pursuing these advancements is the state of organizational data readiness. Despite a growing interest in AI, many companies find that their data is either inaccessible or insufficiently organized for effective AI implementations. Experts emphasize the need for businesses to develop comprehensive data governance strategies to overcome these challenges and facilitate robust AI deployment.
To successfully navigate the complexities of AI deployment, organizations must prioritize data readiness as a foundational element. The journey towards effective AI integration demands meticulous attention to data governance, ongoing evaluation of data quality, and the establishment of clear communication channels between technical teams and domain experts. As businesses continue to explore AI opportunities, ethical considerations regarding data usage must also inform their strategies, ensuring responsible and effective deployments of AI technologies.
Original Source: www.cnbc.com
Leave a Reply