Generative AI tools like ChatGPT, CoPilot, Gemini, and Claude have become powerful assistants for actuaries across the industry. Whether you’re studying for exams, summarizing lengthy material, or drafting communications, these tools can help you think faster and work smarter. 

To get valuable results, you must ask the right questions. Knowing how to write effective prompts will make all the difference in turning generic answers into outputs that elevate you in the workplace and beyond. 

1. Be clear about your goal 

Before you start typing, pause to develop a clear idea of what you want the AI to do. The more specific your objective, the better the response. 

Before: “Evaluate Company XYZ’s financial strength.” 

After: “Evaluate Company XYZ’s financial strength using its latest financial reports. Provide insights on combined ratio, growth rate, and market share.” 

The first prompt here provides an overall task, but it lacks specific action. By going a step further and providing the details of what you expect to see in the output, AI can produce targeted and purposeful responses. 

2. Provide context and constraints 

This is often the core of a strong prompt. Provide the necessary information that your AI tool needs to complete the task at hand. At the same time, AI performs best with boundaries — guide your response by defining anything that it should or should not include. 

Before: “Summarize this report.” 

After: “Summarize this five-page claims trend report into three bullet points for a five- minute presentation to senior management. Focus on frequency trends rather than severity.” 

Adding context (“claims trend report,” “for a five-minute presentation,” “to senior management”) and constraints (“focus on frequency trends rather than severity”) are the keys to producing relevant results. They prevent AI from veering off-topic or producing a response that doesn’t fit your needs. 

3. Assign a role or perspective 

Assigning the AI a role helps it deliver the level of detail you expect. 

Before: “Explain loss reserving.” 

After: “You are an actuarial analyst explaining the concept of loss reserving to a summer intern with no prior experience. Use basic terminology.” 

By defining the role and audience, you shape the AI’s response. This tactic can be especially effective for practicing dynamic communication — something every actuary needs when presenting technical results to nontechnical business partners. 

4. Specify the output’s tone 

AI can write in almost any style — from formal and technical to conversational or persuasive — but only if you tell it what tone to use. This helps ensure that the output matches your audience and purpose. 

Before: “Write an email explaining upcoming rate changes.” 

After:  “Write a concise, professional email to underwriting managers explaining upcoming rate changes. Keep the tone confident and collaborative, avoiding overly technical language.” 

By defining the tone (e.g., professional, approachable, empathetic, casual, neutral), you help the AI calibrate its voice to fit your needs — whether you’re preparing a presentation for leadership, drafting an email to a business partner, or creating study notes for peers. 

5. Provide examples 

One of the most effective ways to help AI understand what you’re looking for is to show it examples. Including an example in your prompt serves as a benchmark for what you’re expecting to see in the output. 

Before: “Write a short bio summarizing each candidate’s work experience.” 

After:  “Write a short bio summarizing each candidate’s work experience. For example: ‘Chris is an actuarial analyst at Travelers with 1.5 years of experience.’” 

By providing an example like this, you’re showing AI exactly what you want. This technique can help avoid the back and forth of editing your prompt to get the AI to deliver your desired output. 

6. Format your output 

AI isn’t just a text generator — it can help you organize information, too. You can request frameworks, outlines, or tables to make complex information easier to digest. 

Before: “Compare GLMs and machine learning for pricing.” 

After:  “Create a table comparing GLMs, gradient boosting, and neural networks for personal auto pricing, with columns for interpretability, data requirements, and typical use cases.” 

When you ask for structure, the output becomes instantly more practical. This can be used to tailor a response style (list, table, bullets), generate data in a specific structure (“provide the result in CSV format”), or create sections in writing (“include an introduction, body paragraph, and conclusion”). 

7. Use iteration 

It’s OK if your initial prompt doesn’t answer everything. Like writing a draft and revising it, you can also refine your instructions to test different outputs or add information. 

Before: “Write a paragraph about how climate change impacts catastrophe modeling.” 

After (using iteration): 

1. “Write a paragraph about how climate change impacts catastrophe modeling.” 

2. “Can you expand that with an example specific to property insurance?” 

3. “Now make it concise enough for a presentation slide.” 

By iterating, you can almost treat AI as a conversation partner that you’re coaching toward your goal. It also can be helpful to experiment with tweaking details of your initial prompt and seeing how it affects your output. 

Conclusion 

As generative AI becomes part of the actuarial toolkit, those who can communicate clearly with it will have a powerful edge. So next time you open your go-to AI assistant, pause before typing. When you define your goal, provide the appropriate details (context, role, tone, format, examples), and refine through iteration, you are maximizing the benefit of AI in the workplace. 

For more information on the uses and risks of generative AI, refer to this article written by David Idoux, “GenAI: Uses and Risks,” included in the December 2024 issue of Future Fellows.