Data, not infrastructure, must drive your AI strategy 

Whether intentionally or not, companies build walls. Different business units use metrics that may not align with those of others. And, if it’s an international organization, data-sharing regulations can add extra borders between teams, preventing efficient collaboration.

Early in the days of generative AI, I asked a chief information officer (CIO) how many data scientists they had. Most are lucky to have one or two, but he answered 800. He didn’t know exactly what they did though, because they spanned multiple business units that didn’t work together.

We helped them establish an AI Center of Excellence (CoE), where groups share knowledge. The result? Several data scientists discovered they could solve a problem that had stumped them before.

Siloed communication stood in the way of progress. It spoke to an underlying data problem, though. Each data scientist had a treasure trove of information. We helped make that data accessible to everyone else, which made much more possible.

That’s the principle behind data centricity—and it’s the key to harnessing AI.

BENEFITS OF A DATA-CENTRIC MODEL

Every company needs an AI strategy. According to McKinsey & Company, 88% of survey respondents said they use AI for at least one business function. If you consider companies planning to integrate AI, that percentage grows. However, looking at companies with effective AI strategies, it sinks like a stone.

To be successful you need a strong business case and must approach the development of an AI application as if it were a traditional enterprise app instead. But that’s where companies need to diverge from the norm. Instead of treating infrastructure as the foundation of your AI strategy, focus on your data estate and see how you can mold your infrastructure to extract value out of it.

Whether it is in the cloud or a data center, the location of the data isn’t as important as the data itself to determine next steps.

Over the years, a data-centric approach hasn’t always been right. You’d have built around developer experience, hosting, application services, things predicated on data but not necessarily centered on it. Times change, though. Data has since become the new oil.

Large Language Models (LLMs) are incredible tools, but not silver bullets. You can’t just point an LLM at multiple data lakes and extract value. Each data repository may be structured differently, with their own security controls. The idea is to develop a system that enables access to data across these locations while maintaining security and controls for each.

Here’s an example: A company collects data from hundreds of partners. Partner A has its own way of sending it in, relative to Partner B, and so on. The data takes different forms: product lists and bundles, pricing information, etc. Under a process-driven model, you’re unpacking and repackaging the data whenever a client is up to renew their contract with a partner.

With a data-centric model, you enable AI systems to access these locations and extract value without having to normalize the data. Making it available to AI is key to unlocking the value across your data estate.

THE SHIFT TO A DATA-CENTRIC MODEL

More than 100 years ago, sociologist William Ogburn coined the term cultural lag. It basically states that technology matures faster than culture. If you try to transform your whole data center from one operating model to another immediately, you’ll experience cultural lag firsthand.

Most companies already have experienced it for themselves, transitioning from data center to cloud. Whenever there have been any of these transformational shifts, the successful ones have started off with a clean slate—which very few established companies can do—or they:

1. Start small.

2. Prove the value.

3. Accelerate once they do.

If value materializes, rinse and repeat. If it doesn’t, you’ll have gone about it carefully and should be able to retrace your steps and course-correct, maybe with the help of a partner who has been through it before.

For teammates in our solutions integration centers, in marketing, finance, etc., AI is a powerful tool we’ve played catch-up on, too. Having gone through many of the same data problems after adopting AI early, we’ve turned lessons we learned as “client zero” into repeatable processes to help ourselves and create solutions to help drive clients’ digital transformations forward.

“So how do you start?” CIOs must take the lead: Break down walls and build bridges between departments. Find opportunities to start flipping the equation to data first, but don’t forget your processes and that your people are what will make the magic happen. You’ll find it leads to measurable business success leveraging AI, all through your data.

Juan Orlandini is the chief technology officer of North America for Insight Enterprises.