The promise of artificial intelligence is alluring for enterprises, but reality often tells a different story. Many organizations face a stark divide between where their data stands and where it needs to be for AI, particularly generative AI, to function effectively.

“I do reference an article I wrote in InfoWorld which references a survey from the Enterprise Strategy Group,” said David Linthicum, enterprise technology analyst at theCUBE Research. “The report that surveyed over 800 IT decision-makers revealed that more than three in five organizations have notable gaps in AI readiness, particularly in infrastructure and data ecosystem. This is something I’m seeing as well.”

Linthicum discussed this topic and more as part of theCUBE’s ongoing AI Insights and Innovation series. He delineated the several reasons behind the dearth of AI-ready data, ranging from data silos to poor data hygiene, faulty semantic frameworks and technical debt.

What’s behind the gap in AI-ready data?

One of the primary challenges facing enterprises as they transition to AI systems is their current data state. Many companies have accumulated vast stores of data over years, but these datasets are often messy, siloed and outdated. This disconnect between data reality and AI aspirations forms a critical challenge for CIOs tasked by their boards to build strategic AI systems that can deliver business value, according to Linthicum.

“Companies have legacy data stores, they have huge data complexity problems, they have huge data silo problems,” he said. “Moving to AI or leveraging that data for training data to train their AI systems to know what they need to know to carry out the processing that they want them to carry out is not going to be an easy pie.”

Another obstacle to AI-ready data is siloing — when different departments store their data separately, often with little communication or integration between them. This fragmented ownership complicated the holistic data environments necessary for effective AI deployment. Without breaking down these silos, enterprises can’t hope to develop a comprehensive understanding of their operations, making AI-powered decision-making nearly impossible, Linthicum added.

Third in the list of problems is poor data hygiene. Inaccurate or incomplete data hampers day-to-day operations and undermines AI’s ability to learn effectively. When AI systems are trained on bad data, they produce unreliable insights, perpetuating errors and inefficiencies. Many organizations struggle with outdated or incorrect customer data, inventory records or transaction histories, according to Linthicum.

“Unless you have accurate and reliable data, the AI system isn’t able to train itself to reach a level of understanding where it’s going to be of use,” he said. “I always say bad data in, bad data out. It’s been a mantra for as soon as I started my career. Garbage in, garbage out, and this is no different.”

Finally, companies must develop a common semantic framework for their data. They must invest in data governance practices that establish uniform definitions and interpretations across the enterprise. A common metadata layer that can standardize how different departments understand and process data is vital to creating an AI-ready organization, Linthicum concluded.

“Go into a room of data managers for large enterprises, and ask them to raise their hand if they have a single source of truth for customers, for products and for anything else that should be important to them,” he said. “If they’re honest with you, their hands aren’t going to go up because they don’t have a single source of truth. The data is all over the place and has different meanings.”

Here’s the complete episode:

Image: KENGKAT from Getty Images