Stop Waiting, Start Crawling: How HHS Agencies Can Take the First Step on AI

At health and human services agencies across the country, there are conversations happening about AI that include phrases like these:

• “We know we need to do something with AI.”

• “We’re working on our AI policy.”

• “We’re waiting to see how other agencies use AI first.”

We understand that instinct to be cautious. AI is moving fast, the policy landscape is complicated, and nobody wants to make a costly mistake. But the workforce crisis isn't waiting; agency caseloads aren't decreasing, and the new federal requirements aren't coming any slower.

The Problem You Can't Hire Your Way Out Of

County agencies have spent years doing what they've always done when things get hard: finding a way to make it work. The strategies are familiar:

• Mandatory overtime and extended shifts

• Bringing retired workers back part-time

• Moving staff from one division to cover another

That approach has run its course and has left many agencies with overwhelmed, burnt-out staff members.

The volume of work coming at agencies right now—increased SNAP certifications and redeterminations, new work requirement tracking under H.R. 1, rising caseloads with no corresponding increase in staff—is not a problem you can solve by hiring more people. Even if you could hire them, if your budget allowed it, and if you had the time to train them, they would still be working within the same manual, administrative-heavy processes that are already overwhelming your current team.

You can't use human resources to solve this problem. Consider that the best path forward is doing things differently.

Moving On from AI 101—Are You Ready?

While many agencies are still debating whether to engage with AI, a growing number of their peers have already moved. They've implemented solutions, seen results, and they're preparing for what’s to come, while others stay in the conversation phase.

At TipCo, we spent much of the last year giving AI 101 presentations, discussing what AI is, what it isn't, and whether it's safe. Those conversations are important and we’re glad to have them when needed, but we’ve noticed a shift. There are agencies that are further along in their use of AI because they aren’t waiting for perfect clarity before diving in. That means starting small, learning about AI capabilities fast, and building confidence, trust, and data to do more.

The agencies still in the 101-phase risk falling into a pattern of analysis paralysis, so focused on every possible risk and open question that no action is taken. Meanwhile, the challenges their staff is facing keep compounding.

The first move isn't to run right into AI usage; it's to crawl. Start with one low-risk, high-value use case and see what happens, learn from it, and then do more…start to walk.

The Policy Landscape Is Messy, But That's Not an Excuse to Wait

One of the most common reasons agencies cite for holding back is the policy uncertainty around AI. And it's a legitimate concern. The current landscape is a patchwork:

• Some counties have no AI policy at all

• Some agencies have policies so restrictive that they prohibit meaningful innovation

• Some state-level frameworks were written by people who don't fully understand the technology

• Many policies finalized six months ago are already out of date
  

All of that is not ideal, but it’s not a reason to stop. AI policy in government will never be finished. Because technology is continually evolving, policy will need to shift as use cases emerge and as sentiments and trust in AI usage change. Waiting for the policy environment to stabilize before doing anything means never getting started at all.

So, with that in mind, what can you do? Figure out how to engage with AI in your current policy landscape—at the county, agency, and state level. Determine whether there are use cases that are permissible within the framework and get started there. We’ve seen agencies have productive conversations with regulatory partners because they come in equipped with real experience and data about how AI could be implemented in their organization.

What "Start Crawling" Actually Looks Like

What does a first step look like for an agency that's ready to move forward with AI?

The best starting point is almost always a use case that meets three criteria:

• It involves publicly available information, not client data and nothing sensitive

• It removes a repetitive burden from workers, freeing up time for more meaningful work

• It's easy to measure, so you can show impact quickly
  

Answering routine incoming phone calls checks all three boxes. When EVA answers a call about office hours, case status basics, or documentation requirements using publicly available data, nothing sensitive is at stake, the worker's time is freed up, and the impact is immediately visible in call volume numbers and staff capacity.

That's crawling. It's not flashy and it doesn’t solve everything. But it builds trust—with your IT department, your workers, your state partners, and your clients—that makes the next step possible.

From there, working up to walking gets interesting:

• Policy bots help workers find answers in thousands of pages of documentation in seconds

• Interview tools save 40 to 45 minutes per engagement and double the number of cases a worker can process in a day

• Outbound notification systems reach clients, removing the need for them to call in
  

None of that is available to agencies still waiting to take the first step.

The Window Is Now

The agencies that moved early on AI are starting to show results that are worth paying attention to. In the second half of 2025 alone:

• EVA answered 404,000 calls across TipCo's customer base, giving back 7,713 hours of worker time

• EVA completed 33,000 rights and responsibilities recitations, saving an additional 4,400 hours

Those hours aren't abstract; they're the difference between a caseworker who is overwhelmed and one who has the capacity to do their job well.

H.R. 1 has made SNAP error rates a direct budget liability for states. The volume of work required to implement expanded work requirements is real and significant. The staffing pipeline is not going to fill itself.

The agencies best positioned to navigate what's coming are those already building capacity, building trust in AI tools, and developing the institutional knowledge to use them well. That process takes time, which is exactly why starting now matters.

You don't have to run. You don't have to solve everything at once. But you do have to crawl.

Key Takeaways

• The HHS workforce crisis cannot be solved through hiring alone. Agencies must find ways to do more with existing staff, and AI is the most practical path to getting there.

• Too many agencies are stuck in analysis paralysis, debating AI in the abstract while peers who have moved forward are already seeing real results.

• AI policy in government will never be "finished." Waiting for perfect policy before acting means waiting indefinitely while challenges compound.

• The best first step is a low-risk, high-value use case involving publicly available information; answering routine phone calls is one of the most effective starting points.

• In the second half of 2025, EVA gave back over 12,000 combined hours to agencies, proof that starting small creates real, measurable impact.

Ready to Take the First Step?

You don't need to have all the answers before you start. You just need to start.

Contact the TipCo team today to talk through what a first step might look like for your agency. We're happy to walk through the options—no pressure, no pitch, just a conversation about the problem you're trying to solve.