The State Department quietly released a generative artificial intelligence playbook this month, offering agencies a repeatable, but adaptable, framework for leveraging the technology in their operations.
The playbook includes a high-level generative AI execution checklist encouraging agencies to first develop a foundational data and AI governance strategy before launching small AI pilots and scaling the ones that show value. After developing a governance strategy, agencies should create a dedicated innovation hub and implementation model that helps officials develop technology and policy as they learn through experimentation, the playbook reads.
It adds that agencies should create a shared enterprise AI platform that can house multiple products, not just one. State Department officials did just that by building an enterprise AI platform that includes StateChat, an internal chatbot for sensitive but not classified information, and can incorporate more software. The Energy Department has similarly included an internal chatbot in a broader AI platform known as Joulix.
After setting ground rules for generative AI governance, agencies should forge internal and external partnerships, collaborate with cybersecurity teams and ship minimum viable products. Those steps will help agencies prove what is possible with generative AI, the playbook states.
As agencies roll out generative AI, they should gradually expand access to users, open channels like office hours to field user feedback and define when exactly beta testing ends, the document recommends.
“A core principle of the Department of State’s approach to scaling adoption of StateChat and its generative AI tools was to ‘meet users where they are.’ This meant the team intentionally designed tailored resources, training, and engagements to fit users’ diverse skill levels, workflows, and contexts — rather than requiring users to adapt to a one-size-fits-all approach,” the playbook reads.
Agencies should build a launch plan that includes multiple phases and receive buy-in from agency leadership, the playbook continues. After a product has launched, agencies should allow users to create new AI solutions, build multiple functionalities within a single interface and measure impact through surveys and focus groups.
The State Department isn’t the first agency to put out an AI playbook for the public sector. The Department of Homeland Security put out its own generative AI-focused playbook in 2025 under the Biden administration.
