As businesses increasingly turn to AI and its subset generative AI (GenAI) to drive innovation and gain a competitive edge, many are struggling to move their AI initiatives from successful pilots to full-scale production. According to Tamas Foldi, SVP, Data at HCLTech, the key challenge lies in getting data estates ready to support the demands of advanced AI models.
Speaking at the World Summit AI in Amsterdam, he says: “There is a fundamental shift happening, where companies are moving from traditional data-driven economies to knowledge economies. You need totally different systems to store knowledge, not just data.”
The limitations of traditional data platforms
The rise of generative AI models like OpenAI’s ChatGPT has exposed the limitations of traditional cloud data platforms, which are often ill-equipped to handle the diverse data types required to power these advanced AI systems.
Foldi explains that companies are now seeking solutions that can ingest and connect structured data, unstructured documents, images and other multimedia in a unified knowledge graph.
“These systems are not designed to store images, data streams, unstructured data, handwritten documentation,” he says. “Now what we see from our clients is that they want to know the outcomes and use cases, which require all the assets, all the knowledge, from their enterprises.”
Leveraging knowledge graphs for AI-driven insights
One use case Foldi provides is a professional services firm that was looking to quickly identify the right talent to staff a client project. Leveraging a knowledge graph that connected employee profiles, project requirements and past experiences, the company was able to use conversational AI to get fast, informed answers.
Building this kind of data-driven AI foundation is no easy feat. Foldi stresses that companies must prioritize establishing robust data governance processes to ensure the quality, security and ethical use of their data assets.
“Data quality and data governance are extremely important. But what degree you need to invest into these will be highly dependent on the use case you are going after.”
Foldi points to one client who spent six months just working to eliminate hallucinations and build the necessary guardrails for a public-facing generative AI application. Failure to data governance rights can not only undermine the performance of AI models but also expose companies to significant brand and regulatory risks.
Balancing data governance and innovation
While there's no one-size-fits-all approach, Foldi emphasizes that companies must take a holistic view of their data and AI strategies. By aligning their use cases, data management practices and governance frameworks, organizations can lay the foundation to scale AI from successful pilots to mission-critical production systems.
“It's always a question of how you calibrate your processes,” he advises. “It's an ongoing effort, and especially with AI when you add the data and AI governance into the equation, it’s a challenge to find the sweet spot between the speed of innovation versus ethical principles or strategy goals versus company values.”
