Data infrastructure: The missing link in successful AI adoption

You know the old saying that you can lead a horse to water, but not make it drink? 

The same sort of logic can be applied to AI adoption by modern businesses: You can roll out AI systems, but you can’t force them to use the data they need to operate effectively. 

In fact, without a modern data infrastructure, you can’t feed relevant data into AI systems very well at all — hence why challenges related to data infrastructure modernization are among the top barriers to successful AI adoption for companies across a range of industries, according to a recent Indicium survey about data infrastructure and AI adoption. 

Keep reading for a dive into what the survey reveals about the role of data in AI rollouts, along with guidance on how businesses can solve obstacles related to data infrastructure as part of their AI strategies. 

The inextricable link between AI, data and data infrastructure 

Data has long been important for businesses. But in the age of AI, it has become absolutely critical. 

The reason why is simple: Without data, AI tools and services can do very little. AI can’t identify relevant trends and patterns, summarize information or generate novel content without being able to parse large amounts of information. 

To be clear, we’re not talking here primarily about the generic data used to build AI models, which are usually pre-trained on vast amounts of publicly available information. In the context of enterprise AI adoption, the most important type of data is information that is specific to individual businesses. The ability to feed this type of data into AI solutions is what makes these tools capable of delivering unique business insights, accelerating business processes and so on. Without access to proprietary business data, AI tools can only answer generic questions, not meet the unique challenges faced by a particular organization. 

Ensuring that businesses can connect proprietary data to AI systems is where data infrastructure comes in. Data infrastructure consists of the tools and technology that an organization uses to store, process and manage its data. Maintaining an efficient, scalable data infrastructure — and one capable of accommodating all types of data, including structured as well as unstructured data sources — is absolutely crucial for ensuring that AI tools and applications can connect to the data they need to operate. 

How outdated data infrastructure hinders AI adoption 

Unfortunately, the data infrastructures that many businesses have built over the past decade or two were designed for the pre-AI age, and they fall short when it comes to powering AI tools and services. 

Conventional data platforms are typically slower to develop, and they lack robust built-in data governance and quality features. What’s more, traditional solutions are often designed only to support structured data, making it challenging to feed other types of information — like documents and images — into AI systems. And they may involve multiple disparate parts, impeding efforts to move data quickly and cost-effectively between the various places where it is stored and into the AI tools that need it. 

The Indicium survey findings reflect the inadequacy of traditional data platforms for the AI era. Asked how prepared they are to use data in conjunction with AI apps and systems, nearly half of respondents reported moderate-to-low levels of confidence. 

What’s more, the survey found that preparing data for use with AI tools and apps is the number one reason why businesses are pursuing data modernization projects — highlighting the priority that organizations place on being able to transform data using a methodology that provides scalability and closes the gap between the business and its technology. Other goals, like reducing storage costs, improving data security and speeding up processes, were much less likely than AI to be the driving force behind data modernization today. 

Bringing data infrastructure up to speed with AI 

What, specifically, are businesses actually doing to address data infrastructure challenges? The survey provides some clear insights. 

Common strategies included migrating from on-prem to cloud-based data platforms, a step taken by 80.9 percent of companies that have pursued data modernization projects. Deploying modern data warehouses, such as Snowflake, Databricks, Redshift and BigQuery, is also a widespread data modernization tactic, embraced by 53.9 percent of survey respondents. 

It’s important to note, however, that simply deploying modern data platforms is only one step in data modernization. Equally important is establishing a firm data management methodology and accompanying organizational culture that defines how to meet data governance, quality and scalability needs with assistance from modern tools. Simply moving to newer solutions does not automatically modernize data management processes. 

Notably, implementing data platforms that specifically target AI-centric data management, such as Vertex and SageMaker, was a less common practice, with only 29.1 percent of companies reporting the use of solutions like this. This is likely because, rather than investing in AI platforms alone, businesses are opting for holistic data modernization strategies that can help not just with integrating data into AI-powered applications and tools, but also with enhancing data governance, security and scalability across the board, not just in the context of AI. 

For the vast majority of businesses, investments like these paid off. Asked whether data modernization projects had left them in a better position to use data with AI tools and applications, 95.1 percent of organizations said they had. 

Data modernization as the key to AI success 

The core takeaway is clear: Building better data infrastructure is an essential step in taking full advantage of AI technology. A business can deploy all of the AI tools and services it wants. But without a modern data platform and management methodology capable of ensuring governance, quality and speed, it’s unlikely to achieve much value. 

The good news is that, for organizations whose data infrastructures are currently out of date, coming up to speed with the AI era is far from impossible. It just requires making investments in modern data platforms and methodologies and management cultures that centralize and scale the way companies work with information. 

This article is published as part of the Foundry Expert Contributor Network.
Want to join?