As artificial intelligence (AI) reshapes the Federal landscape, agencies are under pressure to act with urgency to scale AI for broader mission impact. MeriTalk’s 2025 Federal CAIO Outlook study shares insights from 10 Federal CAIOs on how they are driving scalable, secure, and mission-aligned AI adoption.
MeriTalk recently sat down with CT Thomas, AI and data systems technical director at Dell Technologies, and Shane Shaneman, senior AI strategist for NVIDIA Federal, to discuss the report findings and the path to large-scale adoption.
MeriTalk: In the 2025 Federal CAIO Outlook study, 100 percent of respondents said they believe the benefits of AI outweigh the risks, and 85 percent predict that AI will transform their agencies by 2030 in ways they haven’t yet imagined. What excites you most about AI’s potential in the Federal government over the next five years?
Thomas: AI is one of those few inflection points within technology that impacts how organizations operate. We’re at a point where agencies are looking at how they can operate more efficiently and effectively from a mission perspective and from a technology perspective. I’m starting to see organizations really looking at their processes and examining how they can be revamped – with a lens of excitement and opportunity versus staying with the status quo. It’s exciting.
Shaneman: We’ve seen a tectonic shift from software that has been largely focused on retrieval-based operations to software that is much more generative in nature. Generative AI (GenAI) can be integrated directly into any workflow to empower the workforce at every agency, to enhance productivity, and to improve government operations.
This is going to drive powerful efficiencies, because with GenAI, the programming language is the human language. Now employees and constituents can interface directly with an agency’s data to harness the power, the insights, and the intelligence from it. They can do it without having to wait months for new applications to be developed.
MeriTalk: While Federal AI use cases doubled in the past year, most agencies’ 2025 priorities remain focused on early-stage efforts like governance and security instead of scaling AI for real mission impact. How are leading agencies using AI to drive measurable improvements in their mission delivery today? Can you share a few success stories?
Shaneman: The Federal government has been collecting and storing data for decades, which has resided on servers, waiting for humans to retrieve it and use their cognitive load to assimilate data into intelligence. Now we’re changing course and leveraging GenAI to make use of retrieval-augmented generation with large language models (LLMs) and government data to tap into tremendous insights.
For example, the Department of State’s StateChat is serving 75,000 employees around the world by ingesting news reports across 90 languages to power more focused insights and intelligence, helping diplomats make better decisions.
Thomas: These agencies and other AI leaders in government are consulting with other agencies – sharing lessons learned, how they measure success, and how they prioritize use cases. They are helping to advance AI adoption across government.
MeriTalk: Which AI applications do you see as most promising for improving government effectiveness and expediting service delivery?
Thomas: Process automation and workflow optimization will revolutionize service delivery. People who know how to effectively use AI in their day-to-day tasks will be more effective and make a larger impact. Risk management is another important application. This applies to everything from operational processes to contract negotiation and vendor evaluation. AI systems can perform up to 90 percent of the work and provide bias-free, data-driven outputs. And applying AI to cybersecurity and threat detection is going to help organizations maintain a consistent cyber posture at all times.
Shaneman: Data is the foundation for AI, and agencies have a tremendous amount of work to do with their data to get it to a point where they can leverage AI tools and data analytics to produce intelligence from that data. CIOs can work with partners on data management and prioritize their time for decisions around developing applications and policies for leveraging this information.
MeriTalk: CT, you noted that AI leaders in government are helping other agencies along the AI journey. How important are collaboration and partnerships to accelerating AI adoption, and what do agency AI leaders need to keep in mind when establishing cross-agency collaboration and other partnerships?
Thomas: Establishing a coalition across industry, academia, and government is going to be the force multiplier. With AI, I’ve yet to see someone that tried to go fast – alone – and deliver tangible results. We’re starting to see coalitions collaborating across the board, bringing data governance and cybersecurity expertise to the table for AI projects.
Strategic partnerships that share processes, frameworks, and learnings from other projects enable success from the start by providing resources to help agencies scale AI. The Dell AI Factory with NVIDIA is one example. It’s a platform that helps agencies seamlessly develop, scale, and secure AI implementations through solutions and services tailored for AI workloads.
Shaneman: First, agencies should look at how they want to transform their organizations and workflows. And once they understand this, they can build partnerships and teams to help make that vision a reality. For organizations that are struggling with understanding the potential impact, partners like Dell Technologies and NVIDIA can demonstrate the art of the possible.
MeriTalk: Balancing innovation and risk are key themes of the 2025 Federal CAIO Outlook report. What are the biggest risks Federal CAIOs are facing as they work to make AI work for government – and what happens if they fail to address them?
Shaneman: The pace of AI innovation is not slowing down – it’s never going to slow down. Not moving fast enough is a serious risk. Time and time again, because of the accelerating pace of innovation, leaders don’t know if they should invest in the latest innovation or wait for the next. But agencies that put off adoption will find it increasingly difficult to catch up.
The key is to start getting your hands dirty with pilots or proofs of concept. That initial exposure to AI significantly reduces the risk, because agencies can prove the return on investment, helping to justify the prioritization of and investment in AI infrastructure. Agencies can also use this stage to explore privacy, confidentiality, and cybersecurity implications and work through those with an integrated team.
Thomas: The data element is critical. We talk about policy and security and governance – you have to build that around the actual data architecture itself, which hasn’t been a priority for every organization.
Shaneman: We’re going to hear a lot about agentic workflows, which can dramatically improve efficiency. Federal agencies that haven’t prioritized data architecture and started to look at how to make use of LLMs as an interface to their data will not have the foundation to leverage agentic AI.
MeriTalk: Scaling AI remains a major challenge – two-thirds of CAIOs say their agency lacks the infrastructure, funding, and talent to scale AI. What advice would you offer CAIOs as they transition from AI experimentation to full-scale mission integration in 2025?
Shaneman: Agencies are investing in AI capabilities to empower their workforces, but we’ve seen many projects where the organizations didn’t have the necessary compute infrastructure, and they had to throttle back the new user implementations because they didn’t want performance to fall below user expectations. My advice to CAIOs is ‘Plan for success. Anticipate demand and uptake across the organization.’ Planning for success allows them to think about how they need to scale.
Thomas: Shane, you made a great point there. And it’s important for AI implementers to know that planning for adoption across the organization doesn’t mean buying a lot of compute that you don’t need right now. With the Dell AI Factory with NVIDIA, for example, agencies can start with a small infrastructure investment and expand over time without increasing their technical debt.