This article covers best practices for writing AI Interpretation prompts in Brilliant Assessments, including how to structure your prompt, use merge strings, control output formatting with HTML, and request different output formats such as summaries, tables, and action plans.
When you use AI to interpret your assessment results, the quality of your prompt directly impacts the clarity, usefulness, and professionalism of the output. A well-written prompt helps the AI deliver accurate, well-formatted insights that integrate smoothly into your final reports.
Tip: If you prefer to learn by video, see Lesson 12: Building Feedback with AI Interpretations and Lesson 13: Scoring with AI Interpretations.
On this page
Key Tips for Writing AI Prompts
1. Be Clear and Specific
State exactly what you want the AI to do.
- ✔️ Good: "Summarize the strengths and weaknesses shown in the results below."
- ❌ Too vague: "Write something about this."
2. Give Context
Tell the AI what the data means.
Example: "Use the following scores to evaluate the team's leadership capabilities."
Include relevant background if needed.
Example: "Assume the audience is a senior manager with no technical background."
3. Use Merge Strings Thoughtfully
Merge strings insert data directly into your prompt. They pull in respondent answers, scores, tables, and other assessment data so the AI can interpret real results.
Example: "The following answer is about the underlying technology in the product being assessed for investment: {ResponseAnswer.AnswerText[S3 SS3 Q10]}. Please explain any jargon that may be unclear to a non-technical investor and outline potential risks associated with the underlying technology being used."
For the full list of available merge strings and how to set up AI Interpretations, see Creating AI Interpretations for Reports.
4. Specify the Desired Format
Tell the AI how to present the answer.
- "Provide a 3-paragraph summary."
- "List recommendations as bullet points."
- "Write in complete sentences."
5. Set the Tone
If the interpretation is part of a client-facing report, say so.
- "Use a positive and professional tone."
- "Keep the language simple and encouraging."
Format Your Output with HTML, Not Markdown
The AI interpretation output is rendered in a web-based report engine that supports HTML but not markdown. If your prompt asks for bold, italic, lists, or tables, you need to tell the AI to format its response using HTML tags - otherwise formatting characters like **, ##, or - will appear as plain text in your report.
Important: Always include a formatting instruction at the end of your prompt when you want structured output. For example:
"Format your response using HTML tags. Use <strong> for bold, <em> for italic, <ul><li> for bullet points, and <table> for tables. Do not use markdown formatting."
Writing Your Prompt in HTML
The AI Prompt text box supports HTML. If you want your prompt to be well-structured with clear visual separation between your instructions and merge strings, you can write the prompt in HTML and paste the HTML code directly into the text box.
This is especially useful for longer prompts where you want to use headings, bold labels, or lists to organize the instructions clearly for the AI.
Example: Plain text vs. HTML prompt
Plain text prompt:
"The respondent scored 75% in Leadership and 60% in Innovation. Summarize their strengths and areas for development."
HTML prompt (pasted into the text box):
<p>The respondent's scores are:</p>
<ul>
<li>Leadership: {RatingScore.Score[SectionNo=2]}</li>
<li>Innovation: {RatingScore.Score[SectionNo=3]}</li>
</ul>
<p>Summarize their strengths and areas for development.</p>
Both approaches work, but the HTML version gives the AI clearer structure to work with, especially when your prompt includes multiple data points and detailed instructions.
Common Output Formats
You can ask the AI to return its response in different formats depending on what your report needs. Below are some common output shapes, each with an example instruction to add at the end of your prompt.
1. Summary Paragraph:
Best for executive overviews or brief feedback sections.
Add to your prompt: "Write a 2-3 sentence summary paragraph. Format using HTML <p> tags."
2. Strengths and Development Table:
Best for presenting a side-by-side comparison of high and low scoring areas.
Add to your prompt: "Present the results in an HTML table with two columns: Strengths and Areas for Development. Use <table>, <tr>, <th>, and <td> tags. Do not use markdown."
3. Bullet-Point Action Plan:
Best for actionable recommendations or next steps.
Add to your prompt: "List 3-5 specific recommendations as an HTML bullet list using <ul> and <li> tags. Use <strong> to bold each recommendation title. Do not use markdown."
4. Numbered Feedback Report:
Best for structured multi-part feedback with clear sections.
Add to your prompt: "Structure your response as a numbered list using HTML <ol> and <li> tags. Use <strong> for section headings within each item. Do not use markdown."
Example Prompt
The following is a comprehensive example that demonstrates role-setting, merge strings for dynamic data, structured output requirements, HTML formatting instructions, and tone guidance. You can adapt this template for your own assessments.
Example: Leadership Self-Assessment Feedback
<p>You are a professional leadership development consultant providing personalised feedback to a respondent who has completed a leadership self-assessment.</p>
<p><strong>The respondent's section scores are as follows:</strong></p>
<ul>
<li>Leadership: {RatingScore.Score[SectionNo=2]}</li>
<li>Innovation: {RatingScore.Score[SectionNo=3]}</li>
<li>Strategic Focus: {RatingScore.Score[SectionNo=4]}</li>
<li>Ability to Execute: {RatingScore.Score[SectionNo=5]}</li>
<li>Customer Focus: {RatingScore.Score[SectionNo=6]}</li>
</ul>
<p>Based on these scores, provide a professional feedback report for this respondent. Your response must include:</p>
<ol>
<li><strong>Summary paragraph:</strong> A 2-3 sentence overview of their overall leadership profile.</li>
<li><strong>Strengths and development areas table:</strong> A table with two columns - "Strengths" and "Areas for Development". Place competency areas scoring above 70% in the Strengths column and those at 70% or below in the Areas for Development column. Include the score as a percentage next to each competency name.</li>
<li><strong>Top 3 recommendations:</strong> A numbered list of three specific, actionable recommendations based on their lowest-scoring areas.</li>
</ol>
<p><strong>Formatting instructions:</strong> Format your entire response using HTML tags. Use <strong> for bold text, <em> for italics, <table> with <th> and <td> for the table, <ol> and <li> for the numbered list, and <p> for paragraphs. Do not use markdown formatting.</p>
<p><strong>Tone:</strong> Professional, constructive, and encouraging. Write in second person ("You scored..."). Keep the total response under 250 words.</p>
AI Interpretations also work at the cohort level, analysing results across an entire group of respondents. Here is an example of a cohort-level prompt that you can copy and paste directly into your AI prompt textbox.
Example: Cohort Leadership Feedback (Influence Section)
<p>You are an expert organisational development consultant analysing cohort assessment results for a group of leaders who have completed a Leadership Assessment.</p>
<p>Your task is to review the <strong>Influence</strong> section results across all respondents in this cohort and provide a professional cohort-level feedback summary.</p>
<p><strong>Below is a table showing the scores of all respondents in this cohort, broken down by each subsection within the Influence section:</strong></p>
<p>{Tables[SectionNo=2 SubSectionNo=ALL SplitType=R]}</p>
<p>Based on the data above, provide a cohort feedback report that includes:</p>
<ol>
<li><strong>Cohort overview:</strong> A 2-3 sentence summary of how the group performed overall in the Influence section. State the general trend - is the cohort strong, mixed, or underdeveloped in this area?</li>
<li><strong>Strongest subsection:</strong> Identify the subsection where the cohort scored highest on average. Explain what this collective strength means for the organisation and include the average score as a percentage.</li>
<li><strong>Weakest subsection:</strong> Identify the subsection where the cohort scored lowest on average. Explain the organisational risk this presents and include the average score as a percentage.</li>
<li><strong>Score distribution insight:</strong> Comment on whether the cohort scores are consistent (most respondents scoring similarly) or scattered (wide variation between individuals). If there is significant variation, note what this might indicate about differing leadership practices across the group.</li>
<li><strong>Recommendations:</strong> Provide 2-3 specific, actionable recommendations for cohort-level development initiatives (e.g., workshops, coaching, peer learning) that address the weakest areas.</li>
</ol>
<p><strong>Formatting instructions:</strong> Format your entire response using HTML tags. Use <strong> for bold text, <p> for paragraphs, and <ol><li> for the recommendations list. Do not use markdown formatting.</p>
<p><strong>Tone:</strong> Professional, analytical, and constructive. Write as if presenting findings to a senior HR leader or L&D team. Keep the total response under 300 words.</p>
Good to Know
Tips for refining your prompts:
- Test your prompts. Use the Test Interpretation button in the AI Interpretation window to run a few examples and check the results before going live.
- Refine as needed. Small wording changes can significantly improve output quality. Try adjusting the tone, word count limits, or formatting instructions.
- Use the Resolved Prompt view. This shows you exactly what was sent to the AI after merge strings are replaced with real data, helping you confirm the prompt looks correct.
FAQs & Troubleshooting
Use the questions below to troubleshoot common prompt issues.
Comments
0 comments
Please sign in to leave a comment.