Practical Guide to Prompt Engineering for Public Affairs

By Paul Shotton, Co-Founder Advocacy Strategy / Advocacy Academy

I joined the OpenAI Academy this week. a bit late i have to admit... My goal was practical: learn to create custom GPTs for drafting emails, designing quizzes, and navigating policymaking structures.

But amongst the treasures I found I realized that prompt engineering is a skill worth reexploring. It isn't just a technical skill—it's a thinking skill. It forces you to structure tasks clearly, provide context precisely, and define expectations explicitly.

Prompt engineering can strengthen how we plan, write, and communicate. The key is to stop thinking of it as a single command—and start seeing it as a workflow.

Prompt Engineering as a Workflow

When most people first use ChatGPT, they think the goal is writing the perfect prompt. In reality, it's about creating a series of structured prompts that build clarity and precision step by step.

In public affairs, we rarely get useful outputs from one question. Whether drafting a strategy, a SMART objective, or a stakeholder briefing, we need a process that mirrors how we already work:

• Asking questions

• Gathering inputs

• Structuring outputs

• Refining through collaboration

Each step uses the same three-part structure for a well engineered prompt: Task → Context & Role → Expectation.

Understanding the Task–Context & Role–Expectation Framework

Before diving into examples, let's unpack how this three-part structure actually works. It's simple in appearance, but it changes everything about how you use AI professionally.

1️⃣ Task — Define what you want to achieve

This is your starting point. Tell ChatGPT what you need it to do, as clearly and specifically as possible.

Instead of "Help me with the Digital Networks Act," try:

"Draft a one-page briefing on the key provisions of the Digital Networks Act and their implications for telecom operators."

That single shift—from topic to task—moves the model from guessing to performing.

2️⃣ Context & Role — Explain who you are and where you're coming from

This layer provides perspective. Tell ChatGPT who it should be and what situation it's operating in.

In public affairs, context shapes everything—tone, relevance, framing, even terminology.

"You are a public affairs consultant working for a European trade association representing telecom operators. The audience is a senior policymaker at the European Commission."

This allows ChatGPT to adapt its response appropriately, just as you would adjust a briefing depending on whether you're talking to an MEP, a journalist, or a board member.

3️⃣ Expectation — Describe what the output should look like

Finally, set the performance standard. Tell ChatGPT how you want the output delivered and what good looks like.

"Structure the briefing into short sections with headings, use concise professional language, and include three opportunities and three risks for our sector. Ask for missing details before drafting."

This third layer transforms a useful answer into a usable deliverable. It defines structure, tone, depth, and collaboration style.

How These Three Elements Work Together

Each part strengthens the next:

• Task gives focus

• Context & Role adds relevance

• Expectation defines quality and format

Together, they turn ChatGPT from a general-purpose assistant into a specialized collaborator who understands your goals, environment, and standards.

Because public affairs work is iterative, this framework applies at every step of a workflow—from identifying objectives to drafting documents, refining arguments, or preparing briefings.

Design Prompts by Working Backwards from the Output

Here's a mindset shift: don't start by thinking about what you want the AI to do—start by thinking about what you need it to produce.

A good prompt begins with a clear picture of the desired output—its purpose, format, level of detail, and tone—and then works backward to ensure the instruction supports that outcome.

In public affairs, where outputs are highly structured and purpose-driven, this is essential.

For example:

• SMART objective: You want a two- to three-sentence paragraph stating a policy outcome objective that identifies which actors are involved, what policy change is sought, and by when. Your prompt should explicitly describe that format and specificity.

• Stakeholder briefing: You want a one-page document with a biographical section, background on positions and statements, and key asks or discussion points. Your prompt should mention this structure so the AI knows the expected focus and depth.

  
  

The clearer you are about the final product, the more precisely the model can help you build it. The task, context, and expectation should all serve one goal: deliver an output that's not just accurate, but usable.

Example 1: Building a SMART Advocacy Objective

Creating a SMART objective isn't something you can do with a single prompt. It's a structured workflow that unfolds over several steps.

Step 1: Identify your broader goals

Start by defining the big-picture advocacy goals linked to a policy or legislative issue.

Task: "Help me identify the main advocacy goals my organization might have regarding this new EU initiative." Context & Role: "I'm a public affairs consultant advising a European telecomunications trade association." Expectation: "Please ask questions to help me map the key policy themes and opportunities."

At this stage, upload or summarize key documents—internal analysis, briefing notes, or legislative texts—so the model can understand the landscape.

Step 2: Move from goals to policy outcome objectives

Once you've defined your priorities, choose one area to develop into a policy outcome objective.

Task: "Help me formulate a draft policy outcome objective based on our internal analysis." Context & Role: "I'm advising a trade association preparing its advocacy strategy." Expectation: "Ensure the objective reflects the issues identified and is specific about the policy change sought."

Attach or quote from technical documents or internal assessments to ground the objective in evidence.

Step 3: Define the SMART objective

Now turn the outcome into a measurable commitment.

Task: "Help me rewrite this policy outcome objective as a SMART objective." Context & Role: "I'm a public affairs consultant supporting a European telecommunications trade association." Expectation: "Make the objective specific, measurable, achievable, relevant, and time-bound. Ask me any questions needed to define measurement or timelines accurately."

By the end of this sequence, you'll have SMART objectives that are meaningful, achievable, and aligned with strategy—using AI as a co-pilot for structured reasoning.

Example 2: Drafting a Stakeholder Briefing for a Senior Meeting

Now let's tackle a communication example: preparing a stakeholder briefing for a high-level meeting.

Imagine you work for the Telecoms Trade Association. A senior member of your team is meeting with the Director-General of the European Commission to discuss the upcoming Digital Networks Act (DNA).

Your task is to prepare a briefing that's concise, targeted, and politically aware.

Step 1: Clarify the purpose of the briefing

Task: "Help me outline what a stakeholder briefing should include for a meeting between our association and a senior EU official." Context & Role: "I'm a public affairs consultant preparing material for our Director of EU Affairs." Expectation: "Please ask questions about the meeting's purpose, the official's background, and our policy position before suggesting the structure of the briefing."

ChatGPT will then ask for details about the meeting's goal, tone, and desired outcomes—building relevance before writing anything.

Step 2: Gather your source materials

Upload or summarize relevant inputs:

• The official's biography, portfolio, and recent public statements

• Your organization's position on the file

• Legislative summaries, stakeholder analyses, or technical notes

This context is essential for a professional, tailored output.

Step 3: Create the structured prompt

Task: "Draft a stakeholder briefing for our Director of EU Affairs ahead of their meeting with the Director-General about the Digital Networks Act." Context & Role: "You are a senior public affairs professional supporting the Telecoms Trade Association." Expectation: "Structure the briefing into: (1) Meeting objective and background, (2) Profile of the official, (3) Overview of the legislative file, (4) Key messages and talking points, (5) Recommended questions, (6) Risks and sensitivities. Ask for clarification about tone and length before finalizing."

Step 4: Iterate and refine

Ask ChatGPT to:

• Adjust tone (neutral, persuasive, or technical)

• Condense or expand sections

• Add nuance, such as recent developments or quotes from the DG

Through this iterative process, you arrive at a briefing that's not only factually solid but strategically sharp.

From Prompts to Professional Practice

In both examples, AI isn't replacing the professional—it's reinforcing good practice.

Each workflow turns prompt engineering into a form of structured reasoning, ensuring your outputs are:

• Informed by real documents and context

• Organized around clear goals

• Iteratively refined for precision and tone

When done well, prompt engineering strengthens both thinking and execution—making public affairs work more consistent, evidence-based, and efficient.

Final Reflection

I recommend you check out the prompt engineering and other content at the OpenAI Academy. It has changed how I approach my work. It's not about asking the right question once—it's about designing a structured conversation.

For public affairs professionals, that's a familiar rhythm: clarify the task, define the context, and make expectations explicit.