Government grantmakers are navigating real pressure — small teams, growing compliance demands, and decisions that carry real accountability when they go wrong. AI sits at the confluence of all of it, but whether it introduces relief or risk comes down to a solid base: clear policies, clean data, and teams that know how and when to use the tools.
There are concrete steps agencies can take right now that build toward responsible AI adoption without solving it all at once.
Build policies before you build workflows
An AI policy is only useful if it works when it actually matters — when a staff member is facing a real decision and needs to know what to do. That means it can’t stay abstract. It has to be specific enough to guide judgment in the moment, not just set intentions on paper.
Tammy Gibson, Grants Administrator at the Alabama Department of Early Childhood Education, manages roughly $200 million in annual grants and nearly 1,550 pre-K classrooms across the state. Before deploying any tools, her team drew a clear line on where AI would and wouldn’t play a role. Staff could use AI for lower-stakes tasks like improving letters and drafting notifications. Decisions with real consequences, however, stayed with people. The result has been significant. “It’s probably doing the work of 2 or 3 people,” she said of the compliance review process her small team handles each cycle. But the boundaries are non-negotiable. “We don’t make life-changing decisions based on AI. If we’re going to put 18 to 20 children with a teacher, that is not an AI decision.”
Jerry Trotter, a Nonprofit Business Development Specialist with San Francisco’s Office of Economic and Workforce Development, frames the accountability question simply: “Anywhere a decision can harm an organization’s funding, reputation, or standing, that decision has my name on it.” That clarity matters — not just as a guardrail, but as a signal to staff about what AI is and isn’t responsible for.
For agencies building their first policy, the Municipal Research and Services Center maintains a practical AI policy guide and generator for local governments.
Data quality sets the stage for AI rollout
Government grantmakers operate under a level of public accountability that most private funders don’t. Decisions can be audited, questioned, and scrutinized — which means a misleading AI output isn’t just an inconvenience; it’s a liability. Before AI can serve as a reliable tool, the data underneath it must be trustworthy enough to defend.
That means establishing clear data ownership, a single system of record, validation rules that prevent errors at entry, and regular audit cycles to catch what still slips through the cracks. These aren’t just good data hygiene practices; they’re preconditions for getting useful and reliable outputs from AI.
For a deeper dive into data quality practices, check out our guide Data Quality in Grantmaking.
Implementation is a change management challenge
Getting staff to use AI consistently is harder than it looks. The tools are new, people are busy, and uncertainty about what’s expected can slow adoption to a crawl. Some staff will experiment eagerly; others will wait to be told it’s safe. Without intentional effort to bring people along, even a well-designed policy can sit unused.
That means rollout can’t be just a policy memo and a training session. It requires ongoing reinforcement — managers who model the behavior, early wins that build confidence, and space for staff to ask questions without fear of judgment.
Reymon LaChaux put the larger lesson plainly: “Most AI failures aren’t tech failures. They’re change management failures.” His team addressed this directly, moving to role-specific training: prompting and research skills for program staff, governance and oversight for managers, and strategy and risk orientation for executives. The idea is simple — getting AI to stick requires meeting people where they are, not where you wish they were.
If you’re building the foundation for a responsible AI rollout, Foundant can help. Our grants management software is built for grantmakers at every level of government. Data hygiene is built into the platform and helps teams shift from reactive cleanup to proactive quality.
