You've just spent forty minutes copy-pasting data into ChatGPT, asking it to help you write a quarterly performance report. The response comes back as a wall of prose — dense paragraphs, no structure, nothing your VP can skim in thirty seconds. You paste in a follow-up message: "Can you make it look nicer?" The AI gives you something slightly different, but still not quite right. You try again. An hour later, you've burned more time than if you'd just written the report yourself.
This frustrating cycle happens because most people treat AI formatting like a vending machine: push a button, get a result. But formatting is a skill — and like any skill, it responds to deliberate technique. The difference between a usable report and a wall of text isn't the AI's capability; it's the specificity of your instructions. Once you understand how to write prompts that tell an AI exactly what structure you want, you can generate polished, stakeholder-ready documents in minutes instead of hours.
By the end of this lesson, you'll be able to take raw business data — quarterly sales numbers, survey results, project status updates — and prompt an AI to output it in whichever format serves your audience best, whether that's a clean comparison table, a prioritized bullet list, or a tight executive summary that respects your leadership team's time.
What you'll learn:
No prior prompt engineering experience required. You'll need access to any conversational AI tool — ChatGPT, Claude, Gemini, or similar. If you can type a message and read a response, you're ready.
Before we fix the problem, let's understand it.
When you type "summarize this data," the AI is making dozens of quiet decisions on your behalf. Should this be a paragraph or a list? How long should it be? What's the most important thing to emphasize? Should it sound formal or conversational?
The AI isn't being lazy — it's filling in gaps you left open. It defaults to prose because prose is the most statistically common format in its training data. Business reports, academic papers, news articles — they're all mostly paragraphs. So without other instructions, paragraphs are what you get.
Think of it like hiring a contractor to renovate your kitchen. If you say "make it look better," you might come home to blue cabinets when you wanted white. The contractor isn't wrong — they made a choice in the absence of specification. The fix isn't to get a better contractor; it's to give clearer instructions.
Formatting prompts work the same way. The more explicitly you specify structure, length, audience, and purpose, the more the output matches what you actually need. Let's build that specificity, one format type at a time.
Tables are the workhorses of business reporting. They're ideal when you need to compare multiple items across consistent attributes — think product lines vs. revenue, regions vs. performance metrics, or team members vs. project status.
A weak table prompt looks like this:
Put this in a table.
That tells the AI almost nothing. A strong table prompt specifies:
Here's a realistic scenario: you have Q3 sales data for four product lines across three regions. You want to give your sales director a quick comparison.
Weak prompt:
Here's our Q3 data. Make a table.
Product A: North $42K, South $38K, West $55K
Product B: North $61K, South $29K, West $47K
Product C: North $18K, South $22K, West $31K
Product D: North $73K, South $68K, West $84K
Strong prompt:
Using the Q3 sales data below, create a markdown table with the following structure:
- Rows: each product line (Product A, B, C, D)
- Columns: Region (North, South, West), and a final column called "Q3 Total"
that sums each product's revenue across all three regions
- Format dollar amounts with a $ sign and "K" suffix (e.g., $42K)
- Add a final row labeled "Regional Total" that sums each region's column
- Sort rows from highest to lowest Q3 Total
Data:
Product A: North $42K, South $38K, West $55K
Product B: North $61K, South $29K, West $47K
Product C: North $18K, South $22K, West $31K
Product D: North $73K, South $68K, West $84K
The second prompt gives the AI a complete blueprint. It knows the shape of the table, how to calculate derived columns, how to format values, and how to sort. You'll get a table you can drop straight into a slide deck.
Tip: If you're working in a tool that renders markdown (like Notion, Confluence, or GitHub), ask for "markdown table format." If you need to paste into Word or PowerPoint, specify "plain text table" or "CSV format" instead.
Sometimes you want certain columns to stand out — for example, a "Status" column in a project tracker that uses words like "On Track," "At Risk," or "Delayed."
Create a project status table with these columns: Project Name, Owner,
Due Date, Budget Spent (%), and Status. For the Status column, use only
these three values: On Track, At Risk, or Delayed. Sort by Due Date
ascending. Use the following project data: [your data here]
By specifying allowed values for a column, you prevent the AI from inventing its own vocabulary — which matters when a report gets shared across teams who need consistent language.
Bullet points are not just "a list." They're a design choice that signals: these items are parallel, discrete, and worth scanning quickly. The problem is that AI-generated bullets often come out too long, too short, inconsistently worded, or nested in ways that don't serve your reader.
"Parallelism" means each bullet follows the same grammatical structure. Bad parallelism looks like this:
Good parallelism looks like this:
The second list is easier to scan because every bullet follows the same pattern: [metric or topic] + [quantified change or status]. You can enforce this in your prompt.
Prompt template for tight, parallel bullets:
Summarize the following performance data as a bulleted list with these rules:
- Maximum 12 words per bullet
- Each bullet must start with a noun (the metric or topic being described)
- Include specific numbers wherever the data supports it
- Use past tense for completed items, present tense for ongoing ones
- Limit the list to the 5 most significant findings
Data: [your data here]
Sometimes one level of bullets isn't enough. A department update might have three main themes, each with two or three supporting details. That's when you want nested bullets — a parent bullet with child bullets indented beneath it.
Organize the following project update as a two-level bullet list:
- Top-level bullets should be the three main workstreams: Infrastructure,
Product Development, and Customer Success
- Under each workstream, include 2-3 sub-bullets with specific achievements
or blockers from the past sprint
- Keep sub-bullets to one sentence each
Update notes: [your raw notes here]
Warning: Ask for more than two levels of nesting and readability collapses fast. If you find yourself needing three or four levels of indentation, that's a signal to consider a table or a different document structure instead.
Not all bullets serve the same purpose. An informational bullet reports a fact. An action-oriented bullet tells someone what to do. In business reports, the type matters enormously — a status update uses informational bullets, while a recommendations section uses action-oriented ones.
Informational:
Write a bullet-point summary of last quarter's customer feedback.
Each bullet should describe a finding or trend, not a recommendation.
Keep each bullet under 15 words.
Action-oriented:
Based on the customer feedback below, write 4 bullet-point recommendations
for the product team. Each bullet should start with an action verb
(e.g., "Redesign," "Eliminate," "Prioritize") and include a one-line
rationale. Format: [Action]: [What to do] — [Why it matters].
Feedback data: [your data here]
Notice how the action-oriented prompt even specifies the format within the bullet: [Action]: [What] — [Why]. The more specific your format instruction, the less cleanup you'll do afterward.
The executive summary is arguably the most demanding format in business writing — and the one where AI prompts most often go wrong. Why? Because an executive summary isn't just a short version of something longer. It's a strategic document designed for a specific audience (senior decision-makers) with specific needs (fast context, key insights, clear implications).
Executives aren't reading your report to learn about your data. They're reading it to make a decision or form a judgment. Every word in an executive summary should serve that goal.
Before you prompt for one, know what you're asking for:
A prompt that doesn't specify these components will produce something that looks like a summary but reads like an abstract — descriptive rather than actionable.
Let's say you're presenting a customer satisfaction survey to your leadership team. Here's how to build the prompt progressively.
Step 1 — Define the audience and purpose:
Write an executive summary of the Q3 customer satisfaction survey results.
The audience is the executive leadership team. The purpose is to
help them decide whether to approve budget for a new customer success
initiative.
That's a decent start, but it still leaves structure open. Add format constraints:
Step 2 — Add structure and length:
Write an executive summary of the Q3 customer satisfaction survey results.
The audience is the executive leadership team. The purpose is to help them
decide whether to approve budget for a new customer success initiative.
Structure the summary in exactly four paragraphs:
1. Context (2-3 sentences): What was measured, when, and how many respondents
2. Key Findings (3-4 sentences): The most significant results, with specific numbers
3. Business Implications (2-3 sentences): What these results mean for retention
and revenue
4. Recommendation (1-2 sentences): A clear, specific ask for the leadership team
Total length: 150-200 words. Tone: professional and direct, not hedging.
Step 3 — Add the data:
[Same prompt as above]
Survey data:
- 847 customers surveyed in Q3
- Overall satisfaction score: 6.8/10 (down from 7.4 in Q2)
- Top complaint: slow response times (cited by 61% of dissatisfied customers)
- Top praise: product quality (cited by 78% of satisfied customers)
- NPS score: 24 (industry average: 31)
- Customers who cited response time issues had 2.3x higher churn rate
With all three steps combined, you'll receive an executive summary that's sized correctly, structured clearly, and positioned for a decision — not just a data dump.
Tip: The word count constraint is one of the most powerful formatting tools you have. Executives read hundreds of documents. A 150-word summary that respects their time builds credibility; a 600-word summary that doesn't signals that you don't understand your audience.
"Professional" is not a tone — it's a category. Within professional writing, there's a wide spectrum. Compare:
Each of these communicates the same fact with a different posture. For an executive audience considering a budget decision, "direct" usually wins. Specify it explicitly:
Tone: Direct and confident. Avoid hedging language like "may," "might,"
or "it appears." State findings as facts, not possibilities.
Real reports don't use just one format — they combine them. A well-structured status report might open with an executive summary paragraph, follow with a bullet list of key metrics, and close with a table comparing performance across business units. You can prompt for all of this in a single, well-structured request.
Here's a complete prompt template for a business report section:
Using the data below, generate a business report section with the following
structure:
**1. Executive Summary** (100-120 words)
One paragraph. Audience: VP of Operations. Purpose: weekly team awareness
update. Tone: informative and concise. Include the overall project health
status in the first sentence.
**2. Key Highlights** (bullet list)
4-5 bullets. Each bullet: one sentence, starts with a past-tense verb,
includes a specific metric or milestone. These should be wins and progress
updates only — no risks or blockers here.
**3. Risks and Blockers** (bullet list)
3-4 bullets. Each bullet: one sentence describing the risk, followed by
a dash and a one-sentence mitigation or owner. Format: [Risk description]
— [Owner/Mitigation].
**4. Project Status by Team** (table)
Columns: Team, Current Sprint Goal, Status (On Track / At Risk / Delayed),
Notes (one sentence max)
Rows: Engineering, Design, QA, Marketing
Data: [your raw project notes here]
Notice that this prompt uses numbered sections with clear labels, specifies format, length, tone, audience, and purpose for each section separately, and even specifies the exact allowed values for the Status column in the table. This is what "explicit" looks like in practice.
Here's a realistic dataset. Your job is to write prompts that produce three different formatted outputs from the same raw information.
Raw data:
Annual employee engagement survey — 1,240 respondents (92% response rate)
- Overall engagement score: 71/100 (up from 68 last year)
- Top drivers of engagement: career growth opportunities (82% positive),
team culture (79% positive), compensation fairness (61% positive)
- Top drivers of disengagement: unclear career paths (mentioned by 44% of
respondents), limited remote work flexibility (mentioned by 38%),
management communication quality (mentioned by 31%)
- Highest engagement by department: Engineering (81), Marketing (79), Sales (74)
- Lowest engagement by department: Operations (63), Customer Support (65),
Finance (67)
- 23% of respondents indicated they are "actively considering leaving"
(up from 17% last year)
Your tasks:
Write a prompt that produces a comparison table showing each department's engagement score, with a column categorizing each score as "High" (75+), "Medium" (70-74), or "Low" (below 70).
Write a prompt that produces two parallel bullet lists: one for engagement drivers (positive), one for disengagement drivers (concerns), each with 3 bullets that start with a noun and include a percentage.
Write a prompt that produces a 120-150 word executive summary for the CHRO (Chief Human Resources Officer), structured as: one sentence on the overall trend, two sentences on the most critical finding, one sentence recommendation.
Try each prompt in your AI tool of choice. Then refine — if the output isn't quite right, identify which part of your prompt was underspecified and add more detail. That iteration process is the skill.
Mistake 1: Asking for format without providing structure "Make this look more professional" gives the AI no structural guidance. Instead, name the format explicitly: "Reorganize this as a two-column table with headers X and Y."
Mistake 2: Specifying length in vague terms "Keep it short" means different things to different people — and to the AI. "Maximum 150 words" is unambiguous. Use specific numbers.
Mistake 3: Forgetting to specify the audience A report section for a technical team can include jargon and detail. The same section for a board presentation cannot. Always name your audience so the AI calibrates vocabulary and assumed knowledge appropriately.
Mistake 4: Over-nesting bullet points If your bullet prompt results in three or four levels of indentation, simplify. Ask the AI to "flatten to two levels maximum" or consider whether a table serves the content better.
Mistake 5: Accepting the first output without reviewing AI outputs that look polished at a glance sometimes contain fabricated numbers, misattributed findings, or subtle logical errors. Always verify quantitative claims against your source data before sharing any report with stakeholders.
Mistake 6: Ignoring the data format you provide If your input data is messy and inconsistently structured, the AI output will reflect that inconsistency. Clean up your input data before prompting — or include a note like: "The data below uses inconsistent units; normalize everything to thousands of dollars."
Formatting is not decoration — it's communication design. When you tell an AI exactly what shape you want your output to take, you're not being overly controlling. You're doing what a good communicator always does: thinking about your reader before you start writing.
The core techniques from this lesson:
Put them together with clear section labels and you can build a complete, polished business report section in a single prompt pass — with minor cleanup, not major rewrites.
Where to go next:
The prompt is the brief. Write it like one.
Learning Path: Intro to AI & Prompt Engineering