Beyond Buzzwords: The Gritty Reality of Building AI on Bad Data
- By CDOTrends editors
- February 05, 2024
Data quality is fast becoming a party pooper for the ongoing AI party. A recent survey reveals that while the adoption of generative AI is gaining momentum, data quality remains a formidable obstacle.
“Unsurprisingly, generative AI implementation and the data strategies needed to do so successfully continue to dominate bandwidth for most data leaders, regardless of region or vertical,” stated Jitesh Ghai, chief product officer at Informatica. This sentiment underscores a growing concern among data professionals about the readiness of their data infrastructures to support the sophisticated demands of generative AI.
Data quality issues are not just theoretical concerns but are impacting AI initiatives in real time. A staggering 99% of generative AI adopters have encountered various roadblocks, with 42% citing data quality as the primary hurdle. These challenges are compounded by data privacy, governance, and AI ethics concerns. However, despite these hurdles, there is a strong belief in the transformative power of AI. Approximately 73% of data leaders aim to use generative AI to enhance data insights and boost productivity through automation and augmentation.
The survey also highlights a shift in the metrics used to gauge the effectiveness of data strategies. While the previous year focused on utilizing data in decision-making, the current trend leans towards preparing data for AI and analytics. This shift indicates a more targeted approach towards making data AI-ready.
Chris Eldredge, vice president at the data office in Paycor, emphasized the importance of foundational data management: "AI is only as good as the data that trains it, which means for us to be a successful AI and generative AI organization, we must first be a successful data and data management organization—and that will remain a top focus for us in the months ahead.”
Additionally, the survey uncovers the complexities of managing an increasing number of data sources, with 41% of data leaders grappling with over a thousand data sources. This complexity is expected to grow, further challenging data management strategies.
Internal organizational resistance also poses a significant threat to the realization of data strategies. 98% of data leaders admit to facing such obstacles, including a lack of leadership support and difficulties justifying ROI for budgets.
In response to these challenges, data leaders are continuously emphasizing data management. Investments are directed toward ensuring reliable and consistent data quality for AI, with an increased focus on data privacy and protection, data integration, and engineering.
Image credit: iStockphoto/BrianAJackson