Artificial intelligence in government is moving quickly.
Federal agencies are being encouraged to adopt and procure AI more effectively, while public-sector organizations are beginning to pilot tools designed specifically for government services. From chatbots to robotic process automation, agencies are exploring practical ways to make casework more efficient and reduce repetitive administrative work.
This is where the market appears to be heading: not toward replacing caseworkers, but toward helping them navigate information, reduce repetitive administrative work, and spend more time supporting people.
Technology may change week by week, but the goal should remain simple.
AI should return time to caseworkers without removing responsibility or professional judgment.
Where Can AI Help
Tools like ChatGPT, Claude, Microsoft Copilot, and Gemini can help caseworkers get started on lower-risk tasks such as:
- Drafting a general email
- Creating a meeting agenda
- Organizing non-sensitive notes
- Building a checklist
- Simplifying public information
- Brainstorming questions for supervision
- Rewriting information for a different audience
For example…
Create a professional email template reminding a service provider that documentation is due. Use placeholders instead of names, dates, or case numbers. Clearly explain what is needed and the next step.
The result may save time, but it is still only a draft. The worker must review it and decide whether it is accurate, appropriate, and ready to use.
Where AI Should Stop
AI can organize information. It cannot understand the complete circumstances surrounding a child, adult, or family.
General-purpose AI tools should not independently make high-impact decisions involving:
- Safety
- Eligibility or benefits
- Placement
- Credibility
- Services
- Legal status
- Case outcomes
A polished answer is not necessarily a correct answer. AI can misunderstand information, make assumptions, or confidently present something that is inaccurate.
That is why AI should support, not replace, official policy, the system of record, professional judgment, supervisory review, legal guidance, required assessments, or direct conversations with families.
Four Rules For Using AI
1. Use an Approved Tool
A personal ChatGPT or Claude account is not automatically approved for agency work. Different tools and account types may have different privacy, security, and data controls. Workers should always follow their agency’s AI and information-security policies.
2. Protect Case Fnformation
Do not enter names, dates of birth, addresses, case numbers, medical information, allegations, court records, or other identifying details into an unapproved tool. Use fictional examples, broad descriptions, or placeholders when working in a general AI platform.
3. Keep Decisions With People
AI may help prepare questions for a safety consultation. It should not make safety determinations. The greater the potential impact on a person or family, the more important human judgment and oversight become.
4. Verify the Results
Check AI-generated information against a trusted source, such as:
- Official policy
- State or federal guidance
- The system of record
- A supervisor
- Legal Counsel
- An approved internal knowledge source
Never assume the answer is correct because it sounds professional.
A Simple Way to Prompt AI
Caseworkers do not need complicated prompt-engineering skills. A useful prompt usually includes four things:
- Task: What the tool should do?
- Example: Create a checklist.
- Context: Who is it for?
- Example: This is for a caseworker preparing for a provider meeting.
- Guardrails: What should the tool avoid?
- Example: Do not invent policy requirements. Mark missing information as “[Needs confirmation].”
- Format: How should the answer look?
- Example: Keep it concise and organize it into actions before, during, and after the meeting.
A complete prompt could be:
Create a checklist for a caseworker preparing for a provider coordination meeting. Organize it into actions before, during, and after the meeting. Do not add policy requirements that I have not provided. Mark anything requiring supervisor confirmation. Use clear and respectful language.
Useful Prompt Examples
Preparing for supervision
Create a one-page supervision agenda covering workload priorities, upcoming deadlines, cases requiring consultation, and questions for my supervisor. Do not include client information.
Simplifying public information
Rewrite this publicly available program description in plain language. Do not add or remove any requirements. Identify anything that should be confirmed against official policy.
Organizing meeting notes
Organize these non-sensitive notes into decisions, open questions, assigned actions, and deadlines. Use only the information provided. Mark anything unclear as “[Needs clarification].”
Which AI Tool Should I Use?
Tools like ChatGPT, Claude, Microsoft Copilot, Gemini, and Perplexity can help with drafting, organizing, summarizing, research, and brainstorming.
The more important questions are:
- Is the tool approved by the agency?
- Is the information appropriate to enter?
- Is the tool connected to trusted sources?
- Will someone verify the response?
- Is the final decision still being made by a qualified person?
The brand matters less than the environment, the information being used, and the controls surrounding it.
Returning Time Without Removing Responsibility
The most important parts of casework cannot be automated away.
Listening, understanding context, recognizing risk, building trust, applying policy, and making responsible decisions require human expertise.
AI’s role should be to reduce some of the friction surrounding that work. It can help workers begin a draft, organize information, prepare questions, and navigate approved resources.
Used responsibly, AI can provide a faster starting point.
The caseworker must still provide the context, verification, and final judgment.
Additional Resources on AI in Government
Artificial Intelligence at HHS
Human Services and AI: Key Terms & Concepts (APHSA)
National Governors Association Working Group on AI & the Future of Work
Beyond Generation: The Rise of Agentic AI in State Government (NASCIO)
Generative Artificial Intelligence and its Impact on State Government IT Workforces
