AI’s alluring promises have drawn enterprises far and wide to its adoption, hoping for greater operational efficiency, smarter decision-making, and ultimately, more positive business outcomes. AI’s grandiose appeal overshadows its intense demands of data infrastructure; without addressing the ever-present challenges of data—from legacy infrastructure to data silos, data quality issues, poor governance, and more—AI is doomed to fail.
New approaches to data architecture and data management are key toward enabling AI across the enterprise. In DBTA’s latest webinar, Modernizing Your Data Management Strategy for the AI Era, experts discuss emerging technologies and best practices for supporting AI with a robust, end-to-end data strategy.
David Jayatillake, VP of AI at Cube, emphasized the importance of semantic layers in the AI era. Semantic layers, or a framework that communicates what your data means and how to use it, help diverse enterprise teams, tools, and systems comprehend data. However, many semantic layers that exist today are inconsistent, inaccurate, and dry.
“[Semantic layers] are mostly everywhere; they’re in Excel, they’re in your BI tool—so you get inconsistent and ungoverned metrics,” explained Jayatillake. “It would be much better to have [semantic layers in] one place, where you can govern it…and reuse that definition across many different places. This would simplify not only engineering and looking after those definitions, but also access, exploration, and trust in those things.”
Cube’s belief is that through their universal semantic layer technology, enterprises can unify their data infrastructures and business logic to better support new, innovative AI use cases. Powering the next generation of data experiences, Cube Cloud helps end inconsistent models and metrics and deliver trusted data faster to every use case, according to the company.
For Snowplow, the secret to modern data infrastructure is behavioral data—or data describes what people are doing as they do it, according to Yali Sassoon, co-founder and chief technology officer at Snowplow.
“[Behavioral] data is incredible fuel for AI,” and especially in its agentic forms, noted Sassoon. From support agents to personal shoppers, behavioral data’s inherently predictive and explanatory nature fuels these GenAI experiences. However, behavioral data is also inherently challenging to use due to its consistent high-volume, presenting difficulties in its management and governance.
The Snowplow next-gen CDI solves these challenges so that organizations can build GenAI applications powered by their unique behavioral data streams. Offering tooling to track data quality both pre- and post-production; deliver data in real time to cloud warehouses or lakehouses; capture of the full OODA (observing, orienting, deciding, and acting) loop; and provide semantic metadata underpinned by a knowledge graph, Snowplow helps collect, manage, and operationalize AI-ready, first-party and zero-party behavioral data.
Sean Knapp, founder and CEO at Ascend.io, explained that enabling enterprise teams to increase their overall productivity and velocity is a crucial component of data architecture modernization. To achieve this, the Ascend Data Automation Cloud unifies data engineering processes—from building to automation, observation, and optimization—with the most advanced metadata-powered system available, according to Ascend.io.
Teams are struggling to keep up with the demand for data products, which is compounded by the complexities introduced by AI applications, according to Knapp. By offering end-to-end metadata across the lifecycle, Ascend helps enterprises take control of their operations, transforming DataOps with sophistication and visibility to enable successful AI experiences.
Seth Wiesman, director of field engineering at Materialize, emphasized that context is fuel behind AI experiences.
While everyone can agree that context is important, Wiesman posed a necessary question: How can you provide large language models (LLMs) with the most relevant, accurate, and timely context to fuel intelligent, real-time interactions?
With current approaches for enabling real-time workloads being slow, complex, and expensive, enterprises need new solutions to tap into the power of context. Materialize’s real-time data integration platform enables just that, allowing organizations to transform, deliver, and act on fast changing data—just by using SQL.
For example, Materialize’s platform provides enterprises with real-time customer 360s that afford AI models actionable details for unique queries, making that crucial, contextual data fresh, accurate, and rapidly usable.
To view the full, in-depth webinar discussing data infrastructure modernization for the AI era, you can watch an archived version of the webinar here.