Federal agencies are increasingly exploring ways to leverage artificial intelligence (AI) to improve efficiency, enhance services, and drive innovation. MeriTalk recently sat down with Arno Bergstrom, vice president of sales at IT solutions provider CounterTrade Products, to discuss best practices for identifying and prioritizing AI use cases, evaluating IT for AI, selecting data models, and scaling from pilot to broad agency implementation.
MeriTalk: How can Federal agencies evaluate and prioritize potential AI use cases to get an idea of where to start their AI journeys?
Bergstrom: Defining strategic goals and objectives, based on an agency’s mission and its customers, can be extremely helpful identifying if the agency has specific problems that it might be able to solve using big data and AI. This will really help an agency and its technology partners to outline and put a framework around the AI initiatives.
A collaborative approach across all stakeholders is critical so agencies don’t end up investing both time and dollars into a solution that is not ultimately going to complement their mission.
When an organization wants an AI solution, it’s not as easy as going to the supermarket and grabbing a gallon of milk. There are lots of things going into this recipe along the journey. In an open ecosystem, eight or more partners can be involved in the initial stage of an AI solution. They’re all handling different pieces, helping agencies pull in data, analyze it, produce results, and get feedback to make sure it’s providing accurate and reliable information to ensure good decision-making. One of the biggest things that comes up is establishing governance and ethics. We’re talking about handling privacy, transparency, figuring out what data will be used, and making sure that data stays accurate as we train AI systems.
MeriTalk: Many organizations choose to start with proofs of concepts or pilots when beginning their AI journeys. What are the benefits of this approach? What factors do they need to take into account to scale these into production?
Bergstrom: When bringing AI into an agency, I really believe in starting small. AI implementations don’t need to be an astronomical lift or spend. By testing small batches of data, agencies can minimize risks while proving AI’s value. Once members of the test group use it and see how it makes their day easier, they become the biggest champions of AI. The proof of concept or pilot approach demonstrates ROI and operational efficiencies before full-scale deployment.
When an agency is ready to move beyond the pilot phase, it’s important to continue data evaluation and verification. Change management is very important when working with multiple partners on an AI solution. We want to make sure there’s a validation process in place when users provide feedback, so that when they flag an answer as helpful or unhelpful, leadership can trust that this feedback is being properly reviewed and used to improve the AI solution’s accuracy and insights.
In addition, agency IT leaders need to evaluate if their infrastructure – whether it’s hardware, software, or cloud resources – is ready for a full rollout. The data that feeds AI is going to continue to grow, and the infrastructure must be able to support it.
MeriTalk: At Dell Technologies World, Vice Chairman and COO Jeff Clarke said “there is no one size fits all approach” to AI. How do you meet agencies where they are and help them plan for an AI-enabled future?
Bergstrom: Jeff couldn’t have said it better. An agency might be trying to solve a small problem, or they might be trying to deploy a massive solution across the agency. For example, a team that is collecting water science data from edge devices in streams has very different needs than organizations, such as the Department of State or the Peace Corps, that need to provide agency tools to globally dispersed locations.
The approach we take at CounterTrade Products, along with partners like Dell Technologies, is holistic. We look at the problems agencies are trying to solve now with AI, what the next few years might look like, and we help agencies build AI solutions over time – addressing everything from data management to infrastructure to network performance and security.
The process can be overwhelming, so it’s important to create a roadmap of the AI journey and an evaluation matrix for the implementation. This gives stakeholders confidence and the ability to track successes along the way.
When we talk to agencies about their infrastructure, we use assessment tools to understand the systems already in place – computing power, data, storage capacity, and security protocols. Once we understand the environment and the agency’s readiness for AI, we can provide education on resources for rapid and efficient AI deployment.
MeriTalk: How do agencies choose the best-suited models for their specific use cases and data sets?
Bergstrom: Honestly, it can seem like a daunting task. When we look at AI, we need to take a step back to look at what problems we’re trying to solve and how we can leverage data to solve them.
In many cases, we don’t need to recreate the wheel, so we encourage agencies to look at models that are currently in place. We’ve been able to leverage models from one agency’s use cases and pilots to inform models for another agency. Cross-agency collaboration can really help catapult AI for Federal agencies into the future.
MeriTalk: Are there specific solutions or sets of solutions that are helping Federal agencies accelerate progress?
Bergstrom: Two tools that can help agencies build and maintain AI systems over time are Dell APEX Flex on Demand, which helps customers move to a pay-per-use model for storage solutions that can meet their needs today and give them headroom and capacity to grow without having to replace all their hardware. It’s part of the Dell AI Factory, a backbone of flexible, scalable infrastructure that helps agencies consistently deploy AI at any scale and in any location. The AI factory approach combines servers, workstations, PCs, and storage solutions for AI with a networking portfolio, data management options, and data protection solutions.