To Unleash AI, First Elevate Your IT Infrastructure

A guide to get your government agency’s tech stack ready to leverage native and genAI

At last check, the running count of artificial intelligence deployments by federal agencies stood at more than 700 use cases, from guiding Department of Veterans Affairs medical teams in the operating room during cardiac surgeries to identifying ways to bolster the customer experience at the Internal Revenue Service (no shortage of suggestions to offer there!).

Quantifying artificial intelligence activity at the state level is more difficult. Some states are much further along in their AI journey than others. A recent analysis from NASCIO, an organization representing state chief information officers, and McKinsey & Co., found that 34% of states have built proofs of concept (PoCs) and piloted some generative AI (genAI) models, while just 10% have moved genAI initiatives into production.

Regardless of where a government agency stands on the AI adoption curve, exploring the technology has become a near imperative for public agencies as a means to operate more efficiently and productively, and to provide more value to constituents. However, before they can begin to benefit from their AI initiatives, agencies first have some big questions to answer, chief among them: Where to begin with their AI investments and implementations? And, how well positioned is their IT infrastructure to successfully integrate AI in order to maximize those investments?

In my work supporting government agencies in their digital transformation efforts, those that have intentionally laid the IT groundwork for AI stand to realize better outcomes and a quicker time to value on their AI investments. Here’s a look at some of the key IT components agencies should have in place to set themselves up for long-term success with AI.

Establish a Strong Foundation for AI

1. Evaluate your data and take steps to ensure it’s AI-ready. The quality of insight that AI models produce tends to correlate directly to the quality of the data feeding those models. So among your highest priorities in preparing for AI is to ensure the data flowing into your AI models is fresh, reliable, comprehensive, readily accessible and standardized wherever possible. When data isn’t up to snuff, you risk a “garbage in, garbage out” scenario, where the insight your AI models produce is suspect. Tighter integrations among the systems and software that comprise your IT stack also will enable more comprehensive and relevant data sets to flow into your AI models.
2. Move key IT infrastructure to the cloud. Because AI demands a vast amount of computing power, most AI-driven capabilities reside in the cloud. So shifting your business processes and tech stack to the cloud and shedding on-premise legacy systems and hardware is a vital prerequisite to successfully integrating AI.
3. Identify AI capabilities in your current tech stack. Do like the U.S. government is doing and develop a working inventory of all the AI capabilities and use cases across your organization. There’s a good chance the customer relationship management software or the unified-communications-as-a-service or contact-center-as-a-service (UCaaS; CCaaS) platforms your organization relies upon include AI-driven capabilities, for example. This audit of the software in your tech stack helps identify opportunities where AI can streamline business processes and add value for your employees and customers/constituents.
4. Develop a roadmap. What are your goals in integrating AI into your operations? What are the key steps you’ll need to take to achieve these goals? Here you’re identifying use cases that are technically feasible, map to a well-defined business need or problem, align with your business goals, and project to add enough value to the business to justify the investment.
5. Put in place a set of AI ethics and governance policies, one that people throughout the organization recognize, understand and follow. And be sure to revisit and update those policies to keep pace with evolving AI technology, application and risk. Ensuring the AI solution your organization is using, its underlying model, and the output it produces all are readily explainable should be a critical part of governance, in order to mitigate risks related to legal, security and compliance. It also helps inspire trust in the technology and enables auditability to ensure model performance doesn’t drift or degrade over time.

Building on the Foundation

With these fundamentals in place, it’s time to take a hard look at key components of your IT infrastructure to determine if they need modernizing to support AI:

• If your communications network is outdated, upgrade it. AI requires a strong, resilient, reliable and secure network, with plenty of bandwidth, along with the flexibility to support a remote or hybrid workforce, so they can access AI capabilities wherever they happen to be working. In many cases, a software-defined wide-area network (SD-WAN) fits the bill. The state of Washington, for example, has shifted to SD-WAN as it accelerates its AI activities.
• Get serious about cybersecurity. When you integrate AI-based apps and capabilities into your operations, you’re creating additional surfaces that are vulnerable to cyberattack. Safeguarding them from increasingly sophisticated cyberattacks requires equally sophisticated security strategies. At minimum, consider deploying a next-generation firewall. It’s a solid starting point from which you can add other layers of security, such as secure web gateway (SWG), zero trust network access (ZTNA), and a cloud access security broker (CASB). The endgame: a converged network-plus-cybersecurity solution like Secure Access Service Edge (SASE), which combines these and other protective layers within a single, cloud-native software stack, creating an enterprise-level security fabric that protects potentially vulnerable surfaces out to the network edges.
• Seek out software platforms that offer tools powered by native AI and/or genAI, such as UCaaS and CCaaS. In many cases, the providers of these solutions are adding AI capabilities with each new software release.
• Consider enlisting a third-party expert. A soon-to-be-released report from Forrester Consulting and Windstream found that 59% of U.S.-based organizations lack internal IT capacity, and even among those with adequate IT capacity, 65% lack the specialized internal expertise to successfully execute digital transformation projects. Because of this talent shortfall, many organizations are turning to IT managed service providers (MSPs), handing over responsibilities for assessing, implementing, monitoring, troubleshooting, maintaining and updating aspects of their IT infrastructure to an expert. Because ultimately, for government agencies to take full advantage of the growing number of potential use cases for AI, they’ll need dependable technology partners by their side.

Stephanie Billey is Director, SLED (State, Local and Education) Strategy & Development at Windstream Enterprise, which provides managed cloud communications, networking and security services to businesses and public entities across the U.S. www.windstreamenterprise.com

Photo by Tima Miroshnichenko at Pexels.com