Centers of Excellence (CoEs) have become crucial for success with Artificial Intelligence (AI) initiatives. Ryan Rascop, Insight senior business development manager, recently spoke with the “AI in Business Podcast” by Emerj — hear his insights on how to build a high-performing AI practice with CoEs.
The transcript below is edited and condensed for clarity.
Any business that has the goal of creating a robust, standardized and organized AI practice has an AI CoE because they’ve realized it’s paramount to their success. At Insight, we're seeing clients that run the gamut in the AI space regardless of industry. Some businesses need help with defining potential ideas or AI use cases. Some businesses have those defined but don’t have a clear path to execute. Some have models in production, but they're having some issues with optimizing their models and those practices.
For Insight, it's all about meeting the clients where they're at in their journey — overall, AI CoEs can help businesses realize new ways to build and sustain AI momentum in that journey.
Organizations with disjointed practices, multiple data science teams working independently and a lack of value sharing indicate a need for an AI CoE. Large data science and data engineering teams working on AI projects without alignment and standardization can lead to inefficiencies and increased costs.
A business may be hearing about exciting AI use cases out in the ether, and they know that they need to get involved to start adding value to their organization. But in a lot of cases, a lack of resources, a lack of organization — just a general plan around the entire initiative — can hold them back. We encourage agile pathways that can help get businesses to a point where multiple teams are working together to prepare the business for an AI initiative — in their lines of businesses and their leadership, data science teams and IT.
Many businesses will take a giant leap into AI and find that they may not have the acceleration and excellence that they’re actually looking for. With AI, we’ve found that CoEs help our clients strategize more effectively, identifying what they have, understanding what they're after and what their key business opportunities are, and helping them move further down the path by proving the value to the business.
This often includes building prototypes and Minimally Viable Products (MVPs) with them and accelerating development of their practices to show early ROI. But it also includes the other side of things like managing AI compliance, governance and security to offer them continued success as they go down the path.
From an infrastructure standpoint, we see this a lot. High infrastructure costs that come by surprise and inconsistent practices are all signs of shadow AI. Teams working independently, using disparate infrastructures, often leads to financial and operational challenges.
Improving AI practice maturity is definitely something businesses should consider as they go down these shadow AI paths. You don't want everybody working in their own silos in their own disparate type of organizations in these cases. There are definitely trends in AI infrastructure that are helping to bridge that gap. And I don't want to say band-aid that gap but make it so you don't run into those scenarios where we've got multiple different lines of business or multiple different data science teams working in these disparate infrastructures doing their own thing and not creating a financial responsibility within the organization as well.
Organizations should focus on purpose-built AI infrastructure that allows for scalability and standardization. Having a solid infrastructure in place from the beginning prevents the need for constant replacements and upgrades as the business evolves.
The ultimate goal of a CoE is to drive organizational change and standardize practices, processes and compliance. It's important to choose an AI infrastructure that supports the business and can scale as needed.
A platform that allows your teams and your AI CoE to continue to grow and evolve is key. There's a lot of focus on narrowing down what that purpose-built, end-to-end AI infrastructure is going to be and how it will support the business. And we've got some examples of this with some of our technology partners. For instance, NVIDIA has an H100 stack that's powered by Intel Xeon processors that many of our clients are seeing success with. Rather than constantly ripping and replacing hardware or cloud instances down to meet individual needs, they’re able to scale as the business changes; that will also evolve as they grow. You don't want to start out somewhere and have to completely rip and replace, upgrade and go on to the next big thing — you want to start with a solidified infrastructure practice, where you understand what that is, and the infrastructure offers the best bang for your buck for the organization.
Regardless of what solution stack you land on, starting with a solid infrastructure and standardization avoids the need for major infrastructure overhauls and ensures long-term success across the enterprise.