Toward data-centric AI?

By Edeltraud Leibrock and Manuel Schieler

With corporate spending on generative AI rocketing by a factor of six in 2024 alone, GenAI is clearly no longer confined to the sandbox and is now being factored into serious business strategies. Yet while private consumers adopt such tools as ChatGPT in droves, enterprises are seemingly having a hard time putting GenAI to profitable and efficient use. Based on in-depth interviews with 150 data and AI executives, our new study explores the issues involved and plots a strategic roadmap toward data-centric implementation that delivers the hoped-for but as yet elusive business outcomes.

The age of AI reason?

At an eye-watering pace, the evolution of large language models (LLMs) has long since taken artificial intelligence (AI) beyond being mere 'stochastic parrots' - simple statistical systems that regurgitate patterns from training data. The technology is now learning to “think before responding”. Yet if it is to truly deliver on the hype, this kind of human-like reasoning ability must be fed by drawing on data from multimodal sources, including proprietary and tacit organizational knowledge - the informal expertise embedded in chat histories, internal discussions, and employee interactions. Correspondingly, among the three technological frontiers driving AI advancement, data evolution stands as a fundamental pillar, alongside advances in hardware infrastructure and algorithmic innovation:

A subtle but fundamental shift is therefore taking place: Access to fundamental AI technologies such as LLMs and adequate computing power is already essentially a given, so this alone no longer promises a competitive edge. Instead, enterprises must master the art of creating value by properly feeding their own data into these increasingly commoditized systems.

Harnessing unstructured data — A new challenge

Putting structured data to good use is no longer the bottleneck. While synthetic data creation addresses quantity issues, quality and diversity remain significant challenges. As technological advances now permit the integration of multimodal and unstructured data, existing data management and curation strategies have to be adapted at best, or completely redesigned in some cases.

Our whitepaper highlights the need for systematic approaches that will enable companies to feed this wealth of proprietary knowledge into the advanced AI reasoning capabilities that are now coming on-line.

Reality check for ambitious goals?

This is where today’s exciting “AI for all” narrative becomes a little more sobering: While fully 93% of the executives we surveyed are convinced that systematically implementing AI will improve their data management practices, companies across many major industries—such as healthcare and retail—still lament major difficulties in accessing data of a sufficiently high quality. Also, the task of integrating data from different sources presents varying challenges in different sectors, adding a further layer of complexity that many enterprises have yet to resolve.

While the advent of AI raises all kinds of new challenges for corporate users—from ethical considerations to the sheer lack of suitably qualified people—data issues evidently top the list. Each company wants to streamline its internal processes and deliver excellent services to its customers and target groups. So each must learn the value of its own specific data assets in order to do so. Moreover, ways must also be found to ensure that this data can be integrated efficiently in the LLMs, and that the quality of the data thus provided is genuinely fit for purpose.

Our new study digs deep into the evolving data management landscape as organizations adapt to the unique challenges and opportunities presented by generative AI technologies. Drawing on the expert opinions of 150 data and AI executives, it plainly shows the gap that exists between corporate AI vision and the current state of data infrastructure, data readiness and data management practices. And, as the promise of autonomous agents appears on the horizon, it concludes by spelling out practical strategic steps to help every company close this gap and make sure their GenAI deployment delivers genuine, lasting value.

Register now to access the full study and learn how data management unlocks value in the Generative AI age. Furthermore, you get regular news and updates directly in your inbox.