How Data Center Managers for Federal Agencies Can Create AI-Ready Data Centers

About the authors 

Suri Durvasula, VP of Sales Federal Civilian Agencies, Dell
Brandon Bulldis, Engineer Director - Federal Civilian, WWT
Kevin Pearson, Area VP Federal Civilian, WWT

AI is poised to transform government at the federal, state and local levels. As federal agencies, in particular, actively develop AI policies and explore the value of AI use cases, technologists are on standby. Given the building momentum behind generative artificial intelligence (GenAI), waiting for fully baked strategies is not advised. Data center managers should scrutinize their IT ecosystems now for ways to effectively support the influx of workloads that are based upon artificial intelligence/machine learning (AI/ML). Good starting points include:

• Cleaning up your data
• Building a governance model
• Preparing to secure AI

GenAI will change the federal government. But how?

It is widely agreed that GenAI will fundamentally transform the operations of governments and, for that matter, the world at large.

U.S. federal agencies are among the government entities working diligently to ascertain the implications. To this point, a recent FedScoop poll of federal agency leaders shows a whopping 71 percent said their agencies have already created teams to assess AI's impact and plan to implement AI applications soon.

The federal government has already identified more than 700 use cases for AI and is actively prioritizing how to address them. The top federal civilian use cases center around three main focus areas:

1. Digital assistants: Making the people and teams that produce contracts more productive through statement of work (SOW) development, and using GenAI to help create and build clauses or phrases for use in requests for information (RFIs) and requests for proposals (RFPs) and SOWs.
2. Data insights: Data accessibility and correlations, such as legal requirements and finding keywords.
3. Customer experience: Using AI to better serve citizens. For example, by providing them with better answers to common questions.

While governments and agencies are building AI policies as opposed to taking actions that prepare their organizations for the future, technologists are waiting for the proverbial shoe to drop. However, IT may be missing an opportunity to get ahead of AI, which will inevitably hit like a hurricane.

Instead of waiting for policies to be worked out, technologists should take advantage of this time to do everything they can to prepare for AI. Many things can be done while you wait for the next steps. And the more agile and ready you are to implement AI solutions, the better the outcomes. 

6 activities you can and should be doing right now

1. Assess deficiencies and modernization needs

Many agencies still run legacy systems, operate under massive technical debt, and face capability gaps that are not conducive to effective AI/ML implementations. Now is a great time to address modernization opportunities with your:

• Data
• Network
• Cloud
• Cybersecurity
• Workforce

2. Be an enabler of the AI conversation

Across the government, agencies are adding Chief Artificial Intelligence Officers (CAIOs) — some of whom may have a data science background while others may hail from an operational AI background. The key with this role in this conversation, per the OMB memo, is that the officer is able to break down traditional, siloed AI conversations to support the Mission Outcomes while enabling the IT organization.

AI may very well be the most transformational game changer for the government ever. It's vastly important to make sure your voice is heard, and that all the relevant possibilities impacting IT and the business of government have been considered from the onset. Every aspect of IT must be included and considered in these conversations. After all, chances are that a large portion of your IT budgets today will transform into AI budgets within the next five years.

3. Break down organizational barriers

Even within IT, silos persist. To truly embrace AI, all IT groups need to collaborate and work toward the same goal, including the following teams: networking, data center, cloud computing and beyond. Numerous groups within IT (and in some cases outside of IT) own and impact significant elements of the underlying initiatives required to successfully enable AI, including:

• Data cleansing
• Modernization
• Cloud migration
• Developing an effective zero-trust security approach

If all the accountable groups are not working together and communicating effectively, important AI enablers could fail, and the entire agency could feel the negative impact.

4. Rethink how and where data is accessed

Graphics processing units (GPUs) and storage are fundamentally shifting as a result of GenAI and how people will consume compute in the long term. This means you need to reconsider how people access your data.

With GenAI, you need to rethink how you provide these architectures and how you're providing these services. You may need to take a half-step back to confirm the path you're on, the architecture you're moving toward, makes sense in terms of GenAI's new high-performance needs — or you should prepare to shift to AI-ready architectures.

Most organizations won't necessarily need to go back to the drawing board. Many existing data modernization projects are already in the midst of reconsidering where data resides, how it's backed up and recovered, how it's accessed, and so forth. Obviously, you need to modernize how and where data is accessed even before data rationalization can happen. You can continue with your existing data projects but do so with an AI-forward mindset.

5. Clean up your data

Data governance and data management have always been a challenge for government entities. Now that agencies are multi-modal and data is everywhere — from the cloud and on-premises to colocation facilities and software as a service (SaaS) — it's gotten even more complex. 

In the future, the model of AI you implement — whether a large language model (LLM) or a more systemic form of AI solution that isn't GenAI — will determine systems and other needs. At first, though, the most important thing is ensuring you have clean data and the right governance over it so you don't introduce garbage into a clean environment.

The biggest problems include false AI, misinformation and inadequate guardrails. As the saying goes, "garbage in, garbage out." For any solution powered by AI, the first thing you need to do is look at your data and clean house before the inevitable happens.

Rationalizing your data is a simple thing you can do today without buying any GPU-driven system or building out a big AI capability. You have to know where your data is, have access to it, and make sure it is squeaky clean and accurate. This is an important task that should not be taken lightly.

6. Ensure clean data with data governance

Data governance is pivotal for federal agencies, especially in the context of AI implementation, due to its implications for security, accuracy and regulatory compliance. Here are three reasons why:

• Enhancing national security: Robust data governance frameworks ensure sensitive information, such as national security data processed by AI systems, is accessed only by authorized personnel. This minimizes the risk of sabotage by malicious actors.
• Promoting AI integrity and trust: By establishing clear rules on data handling and processing, data governance helps maintain the accuracy and integrity of AI-driven decisions. This is extremely important for applications like predictive policing or security clearance assessments.
• Ensuring regulatory adherence: Federal IT projects, especially those using AI, must comply with myriad regulations concerning data privacy, security and ethical use. Data governance frameworks aid in navigating these regulations, such as the Federal Risk and Authorization Management Program (FedRAMP), ensuring that AI projects align with federal standards and avoid legal and ethical pitfalls.

Security of AI

Before bringing AI workloads in-house, you also must consider how to secure AI systems and tools and how to best protect your valuable data. That means you have to build out a security posture that allows your data to live under one safe umbrella. You'll need to define guardrails and implement new frameworks as part of this process. AI will comprise the newest workloads in any given data center. 

Protecting AI workloads involves:

• Keeping malicious actors from injecting bad data
• Protecting data as it goes out, making sure AI doesn't go haywire
• Building a cyber fence
• Adding security measures
• Implementing data cyber principles

Cybersecurity is complex and what you've read thus far is relatively high-level. For a step-by-step guide to unlocking the power of AI while protecting your data, follow the link below for a free report from WWT Research.

Count on experienced partners

With Dell and WWT, you can quickly and easily create AI-ready data centers so you're ready for go time.

Dell has made significant investments in developing AI systems, from RAG to foundational models, alongside a vast storage portfolio designed to integrate seamlessly with AI solutions, such as Dell PowerScale. 

WWT can help you build out your whole AI strategy and story, including software and services, in our AI Proving Ground, the first AI testing environment with technology from multiple OEMs. WWT's large data practice includes experts who deeply understand data science and infrastructure. In addition, WWT's Advanced Technology Center (ATC) can be tapped for market research and education.

Discover the future of AI for federal IT. Visit WWT's AI Proving Ground today and schedule a discussion of how we can help you lead your agency into a new era of innovation and efficiency.