Federal Chief AI Officers say the benefits of AI outweigh the risks, the challenge is finding practical use cases that scale that can be scaled across government IT
by Intelliworx
The federal retirement management system (RMS) is getting a boost from artificial intelligence (AI), according to reporting by FedSmith. The processing of retirement packets has long been a weak spot for the federal government.
There’s a backlog of applications, too. Those applications are all still processed by hand. That typically takes months, and in some of the more complicated cases, can take years. Ian Smith, who penned the piece, reports that the average processing time for the nearly 9,000 retirement packets approved in April 2025 was 49 days.
He notes, while AI isn’t being used to digitize old paper records, it is preventing new applications from piling up. The system has reportedly been tested for accuracy and is functioning as intended. The result sounds promising: retirement applications can be processed in under a second.
This is just one of many AI projects the federal government is experimenting with as it strives to modernize its sprawling IT infrastructure. That’s what makes a new report on the government’s efforts to pilot AI projects so timely.
The editorial team at MeriTalk put together the report by collecting input from 10 Chief AI Officers (CAIOs) who were currently working for federal agencies as the data was compiled. They also surveyed these senior tech leaders, and while 10 respondents isn’t a big sample, their expertise, the subject matter and quantification of their answers make the findings informative.
(Click for larger image)
Below are a few findings that stood out to us:
1. The benefits of AI outweigh the risks
The report found “100% of CAIOs feel the benefits of AI outweigh the risks.” This is an important point because there are drawbacks to everything, and that’s not being overlooked with AI. The group collectively named “establishing AI governance and compliance” as the top priority.
That said, AI is an arms race. The efficiency and efficacy to be gained in routine government business processes could well be misused or even weaponized in another context. Either way, the U.S. needs to win this race, and we seem to be doing it with the right guardrails to mitigate risk.
2. Some of the best use cases are yet to be imagined
A majority of CAIOs (85%) “predict that AI will transform their agency’s operations by 2030 in ways they haven’t yet imagined.” However, they uniformly noted “they lack the funding and resources to meet their goals.”
Budget to make things happen was a common thread in the report, yet this is a proverbial ‘chicken and egg’ style scenario. No one is going to write a check with taxpayer money for a project unless they have a clear understanding of the proposal, requirements and potential impact.
The good news is that there are now some 1,700 emerging AI use cases federal agencies are exploring right now to flesh this out. Invariably, word will get out about the most successful experiments. In turn, that will help provide the clarity needed to program a budget for specific projects.
3. Two distinct federal approaches to AI
“Federal CAIOs fall into two distinct camps – those building the foundation and those accelerating mission-driven adoption,” MeriTalk says each of these two cohorts has a different approach to implementing AI.
The first group, dubbed the “builders,” is focused on laying the foundation for AI within their agency. This is the “most common” scenario and their priorities for AI are as follows:
- “Establishing governance”;
- “Implementing security”;
- “Scaling infrastructure and computing”; and
- “Engaging stakeholders and promoting adoption”.
The second group, dubbed the “accelerators,” appears to have its foundation set. They are now focused on scale – scaling use cases, adoption and results. This is “less common” of the two approaches and suggests these agencies have gained a bit more maturity around AI. As such, their priorities are a bit different from the former:
- “Exploring new use cases that balance innovation and risk”;
- “Evaluating current use cases”;
- “Expanding AI talent and training”; and
- “Measuring and reporting outcomes”.
The second cohort has an advantage:
“CAIOs in the accelerator camp are 2x more likely to feel they have the authority to advocate for meaningful change.”
That makes sense given they seem to be further along and have a greater command of the details necessary to make a sound business case for their projects.
4. The top challenges for AI in the federal government
MeriTalk presented these CAIOs with this question: “What are the biggest challenges you expect to face as you implement your 2025 AI goals?”
Here’s how the answers stacked up:
- 100% said “insufficient funding or resources”;
- 83% said “lack of internal AI expertise”;
- 67% said “data quality and accessibility issues”; and
- 50% said “difficulty integrating with legacy systems.”
In practice, leaders will never have ‘enough’ resources. Deciding how to best apply finite resources for maximum benefit is part of the challenge of leadership.
Yet there is a soft skill that could help here and noted in this section of the report: strategic communications. As one CAIO remarked in an open-ended comment:
“It is difficult to do the very important strategic communication part … There’s often not a lot of overlap with someone who has a background in strategic communications, a good technology background, and experience on policy issues.”
Clear communication is the path to soliciting ideas for AI projects, building a business case, fostering user adoption, and reporting on successes.
5. Where leadership lends its support
Leadership support is often an intangible characteristic that can make or break an IT project. In this report, CAIOs noted they had the “most support” for “establishing AI governance and compliance.” They also reported the “least support” for “strengthening interagency and public-private collaboration.”
Strategic communications is another area that would facilitate inter-agency collaboration on AI. The communications process can help identify those projects that can drive efficiency across several organizations, pool development resources and produce scalable results. The federal retirement system is a good example because it will impact virtually every federal employee.
In our reading of the tea leaves, this is also the purpose behind the government’s renewed focus on commercial off-the-shelf (COTS) software. With 430 agencies and sub-agencies in the federal government, each with a distinct mission and needs, the high price and niche utility of custom software may put some AI projects out of financial reach.
What’s the takeaway for agency leaders?
This report makes it clear that agency leaders need to get involved with AI. They need to get their people learning the fundamentals and experimenting with the available AI tools. From our vantage point, the key to that is removing the barriers so your team can begin finding ways to use AI to create value, maintain consistency and ensure quality.
* * *
The full study is freely available here: Tech Tonic 2025 Federal Caio Outlook.
If you enjoyed this post, you might also like:
SaaS case study: How the Department of Veteran Affairs modernized financial disclosure
Image credit: MeriTalk and Unsplash
