Prompt Engineering for Public Health
Our February AI in Public Health Community of Practice session dug into one of the most practical skills public health professionals can develop right now: prompt engineering.
We welcomed AAron Davis from Wichita State University's Center for Public Health Initiatives, who walked us through his team's approach to getting better results from AI tools like ChatGPT, Copilot, Claude, and Gemini.
F&T Labs offers AI training programs built specifically for governmental public health, including a dedicated Prompt Engineering Deep Dive course. We also build PH360, purpose-built AI for public health that keeps your data secure and your workflows compliant.
Don't Just Say "Make It Better"
AAron opened with a metaphor that stuck: walking into a diner and saying "feed me." You will get something, and it probably will not be what you wanted. The same applies to AI. If you drop a document into ChatGPT and type "make this better," you leave the system to guess what "better" means. Good prompting means telling the tool exactly what you need.
He framed most AI use in public health around three activities: assessing something (reviewing a policy draft, comparing two programs), brainstorming (generating creative engagement strategies, identifying questions to ask elected officials), and creating something new (drafting agendas, building checklists, writing SMART objectives).
Prompt Formulas That Work
AAron shared several structured frameworks to help practitioners build more effective prompts.
APE
Action, Purpose, Expectation
Define the task, explain your goal, and state the desired outcome. For example: "Review the script for our upcoming mental health awareness workshop and suggest edits to make it more engaging and easier for high school students to understand. The goal is to increase student comfort in discussing their personal mental health issues. I'd like the final version to feel conversational, informative, and inclusive."
RISE
Role, Input, Steps, Expectation
Assign a role, describe your source materials, request specific steps, and state what you expect. This works well for multi-step tasks like drafting communications or creating structured content.
CREATE
Context, Role, Example, Action, Tone, Expectation
The most comprehensive framework AAron shared. It covers context, role assignment, examples, the specific action you want the AI to take, the tone you want in the output, and your expected format.
The key insight across all of these: you do not need to memorize a formula. The formulas exist to train your brain to provide complete context when you communicate with AI tools.
Roles Change Everything
One of AAron's most compelling examples involved a workshop participant in Texas who was revamping a crisis website for students. The team initially prompted the AI to act as a "UI expert and website developer" to review the site. The results were fine. Then AAron suggested a simple shift: tell the AI to act as a student in crisis. The response changed dramatically, surfacing feedback that closely mirrored what students themselves had been saying.
The lesson is clear. The role you assign shapes the output. A "health policy analyst" will give you a different review than a "Medicaid recipient" will. A "public health communications specialist" writes differently than an "elementary school teacher" does. Experimenting with roles is one of the fastest ways to improve what you get from AI.
Practical Tips from the Session
Ask AI to Ask You Questions First
AAron uses this technique in roughly 70% of his prompts: before the AI generates anything, ask it to pose up to 10 clarifying questions first. The tool will identify gaps in your prompt and give you a chance to fill them in. This one habit can dramatically improve output quality on the first try.
Tell It What You Don't Want
Most people only describe what they want in a response. AAron emphasized including what you do not want: "Do not use jargon." "Do not exceed 200 words." "Do not reference political parties." Constraints sharpen results.
Provide Raw Materials
Upload notes, paste in rough drafts, attach meeting agendas. The more context you give, the more targeted the output. Scattered notes are fine. AI can work with messy input as long as it has something to work with.
"If you wouldn't put it on a billboard, don't put it in your prompt."
AAron also raised an important caution: if you are using a free AI tool, understand where your data goes before you enter anything sensitive. Proprietary information, protected health information, and confidential data should never go into a system without verified privacy protections.
Treat AI Like a Collaborator
AAron's closing message resonated throughout the session: AI performs best when you treat it like a capable intern, not an answer machine. That means providing context the way you would brief a colleague, reviewing and editing outputs the way you would review an intern's draft, and redirecting the conversation when it goes off track.
Good prompting, AAron emphasized, is really good leadership. It requires clear communication, critical thinking, and the willingness to engage with the output rather than copy and paste it.
We are former public health officials, not consultants who theorize from the outside. We use AI in our own practice every day, which means we understand the real constraints departments face: limited budgets, HIPAA obligations, skeptical leadership, and staff who are already stretched thin. Our training programs meet your team where they are and build toward something your department can actually sustain.
We train public health regions, state associations, and individual departments. Every program starts with your tools, your policies, and your specific challenges.
GenAI for Public Health
Gives program staff and epidemiologists the foundation for safe AI use, from epi analysis to report generation, with security protocols built into every lesson.
Prompt Engineering Deep Dive
Teaches advanced techniques for chaining complex tasks and building consistent, high-quality outputs. Participants leave with an advanced prompt library tailored to public health workflows.
Building an AI Policy
Walks leadership teams through developing a comprehensive governance policy in three facilitated sessions. You make all the decisions; we provide the structure.
Annual AI Compliance Training
Keeps your entire department aligned on approved tools, current policies, and incident reporting procedures.
Executive Governance
Helps health directors and CIOs bridge the gap between unauthorized AI use and compliant implementation.
Technical and HIPAA Training
Equips IT and compliance officers to evaluate their technology stack, establish security protocols, and create audit-ready documentation.
We also offer a free AI basics course if you want to start there.
Ready for More Than Training?
PH360 is purpose-built AI for public health. It is HIPAA-compliant, designed for governmental workflows, and keeps your data within a secure environment your department controls. When your team is ready to move beyond general-purpose tools like ChatGPT and Copilot, PH360 provides AI that understands public health from the ground up.
A great first step is making sure your department has an AI policy in place. Our guide on 12 Things Your AI for Public Health Policy Needs walks through each essential component.
Let's Figure Out What Works for Your Department
We are not here to sell you a generic AI solution. We are former public health officials who are actively implementing AI in practice, and we want to understand your constraints before we ever suggest a path forward. Whether you are a regional consortium, a state association, or a single county department, we will meet you where you are.
Schedule a conversation with us and we will work through what makes sense for your team.
Schedule a Meeting