P2P

69 I L T A N E T . O R G implementations stall at the pilot or proof-of- concept phase, having a robust data governance program can help organizations streamline Gen AI implementation and improve its output, tap its transformational benefits faster while minimizing risk, achieve a much greater return on their investments, and achieve their corporate objectives. The argument for enterprise organizations to "ground" Gen AI models on internal or industry data vs. the Internet is that it leads to highly focused outputs and potentially fewer hallucinations. The other side of this argument is that the voracious data requirements of today 's Gen AI models can increase organizational risk if given keys to the proverbial digital kingdom without already having rock-solid data governance policies and initiatives. LEVERAGING DATA GOVERNANCE TO IMPROVE GEN AI OUTPUT How can organizations most effectively leverage Gen AI models while also adhering to data governance best practices? Here are some things to consider: • Data Quality: Ensuring data is clean, accurate, and unbiased is crucial for authoritative Gen AI outputs, but large internal datasets or external industry data can be challenging to clean, validate, and normalize. Using the previous types of large internal or external datasets necessitates additional steps to ensure the data is accurate, reliable, and credible and minimize the chance of hallucinations. For example, if grounding data is biased, Gen AI outputs will reflect those biases, resulting in misleading, discriminatory, or unfair outcomes. Additionally, it is crucial to continuously monitor and update datasets to maintain accuracy and reliability over time. Robust data governance establishes clear data quality standards that ensure consistency, eliminate unnecessary duplication, and streamline data-related processes. It does this by assigning data stewards and domain experts, whose sole responsibility is ensuring data quality, security, and adherence to stated policies. AI tools can help supplement some of this work, but data stewards and data domain experts are needed to provide critical humans-in-the-loop quality control. For example, AI can overhaul the quality assurance process, automating and significantly refining data integrity with minimal direct human involvement. FEATURES

• Cover