With the explosion of AI capabilities and adoption in the past year, a top-ofmind question for leaders across the enterprise is, “How do I use AI to empower my business performance?”. One of the critical mechanisms identified by insightful leaders as a potential opportunity for AI lives within the data space. The augmentation of analytical capabilities or cost management opportunities resulting from AI transforming analyst work is immense. Imagine if every employee at the firm had their very own analyst in their pocket who could ensure that all choices across the organization are wholly data-informed and evidence-based to their core. Such a reality would radically transform most businesses from top to bottom, hugely maturing their decision-making insight and resulting in massive value creation.
Given the potential size of the prize, the next logical question lies in how this AI analytical capability can be developed. While hundreds of new startups and mature data companies are now focusing on solving this problem, a more fundamental solution already exists. Recent research has shown that the implementation of a knowledge graph, or more broadly, a metadata system, can improve LLM-powered analytics by more than a factor of three in terms of the accuracy of analysis. Said differently, metadata, not a shiny new SaaS product, is the key to the future of AI analytics.
What is metadata? In short, metadata refers to data that describes other data. Most commonly, in the data sphere, we think about this as a structured semantic layer that sits on top of the database or data lake and is used to; centrally define metrics, govern business logic, and provide transparency as to the interconnected nature of enterprise data. An easy way to understand metadata is through metaphor; if a piece of data is like a package of cookies at the supermarket, the metadata is the nutritional label on the back of the cookie package that describes different data characteristics. AI currently requires this ‘label’ to reduce inaccuracy from hallucination and better understand what end users are querying for. Having metadata describes a richer data context yields more relevant and accurate results when requesting analytical insights. While the highly mature companies on the cutting edge of data sophistication have already built out this metadata (for example, Google’s knowledge graph powers a vast amount of their products), for most firms, this is a net new investment required to get the best out of AI.
"The AI analyst is coming and will be critically valuable to the enterprise success of many firms"
The companies who have realized this prerogative are quickly shifting focus and investment to capturing this value. The most common approach in the industry is adopting transformation tools (especially debt), which are cloud-agnostic and focused on stitching together data governance, transformation, and control into a single platform. Luckily, semantic layers (the leading approach to metadata management) are core to these platforms. While new platforms require investment, the return on that investment, especially as a result of the power of AI, now makes deep capital allocation towards these tools a requirement for any firm that wants to maintain competitiveness. In the words of Frank Slootman: “There’s no AI strategy without a data strategy. The intelligence we’re all aiming for resides in the data, hence the quality of that underpinning is critical”. In sum, the AI analyst is coming and will be critically valuable to the enterprise success of many firms. Moreover, building that AI strategy on strong metadata foundations is a prerequisite for success, and the leading firms have recognized this precondition for success and are already investing vast capital to capture the opportunity. If data leaders don’t want their firm to fall behind, they must act now and begin building the bridge toward the future to compete and win in an AI-empowered world.
