Analysis and Synthesis

Grok as an Analysis Tool

Text analysis is where Grok shines. Summarizing a 50-page document, extracting structured data from free text, comparing arguments, classifying customer feedback: tasks that take hours manually happen in seconds with the right prompts.

Summarize Effectively

A good summary isn’t just a shorter text. It’s a transformation that captures the essence while eliminating the superfluous. Summary quality depends on how precise your request is.

Executive Summary

Summarize this document in an executive summary of 5 lines maximum.

Document:
[paste the text here]

Instructions:
- Start with the main conclusion
- Include the 2-3 most important figures
- Target audience: CEO who has 30 seconds
- No technical details or methodology

Structured Summary

Summarize this report following this structure:

1. Study objective (1 sentence)
2. Methodology (1 sentence)
3. Key results (3-5 bullet points)
4. Recommendations (2-3 bullet points)
5. Limitations (1-2 bullet points)

Document:
[paste the text here]

Constraints: each section must not exceed 2 lines.

Comparative Summary

I have two articles on the same topic. Summarize each in 3 points, then identify points of agreement and disagreement.

Article 1:
[text]

Article 2:
[text]

Output format:
## Article 1 - Key Points
## Article 2 - Key Points
## Agreements
## Disagreements

Data Extraction

Extraction transforms unstructured text into structured data. This is particularly useful for processing volumes of documents.

Entity Extraction

Extract the following information from this contract:

- Parties (full name, role)
- Signing date
- Contract duration
- Total amount
- Termination conditions
- Any penalties

Contract:
[paste the contract text]

Output format: JSON with the keys above.
If information is not found, indicate "unspecified".

Quantitative Data Extraction

Extract all numerical data from this financial report and organize them in a table:

| Indicator | Value | Period | Variation |

Report:
[paste the text]

Include only explicitly mentioned data, no estimates.

Comparison and Analysis

SWOT Analysis

Perform a SWOT analysis based on this company description.

Description:
[paste the text]

Format:
## Strengths (internal, positive)
## Weaknesses (internal, negative)
## Opportunities (external, positive)
## Threats (external, negative)

For each element, briefly justify with an excerpt from the source text.
Maximum 4 elements per category.

Argument Analysis

Analyze the arguments presented in this text:

Text:
[paste the text]

For each identified argument:
1. Summarize the argument in 1 sentence
2. Evaluate its strength (strong / moderate / weak)
3. Identify supporting evidence or data
4. Identify any weaknesses or biases

Be objective and do not take a position.

Classification

Customer Feedback Classification

Classify each of these customer feedback items by:
- Category: product, delivery, customer service, price, website
- Sentiment: positive, negative, neutral, mixed
- Urgency: high, medium, low

Feedback:
1. "The package arrived open and the product was broken"
2. "Very satisfied with the quality, I recommend it"
3. "The site crashes systematically at payment"
4. "Price is fair for the quality, nothing exceptional"
5. "Impossible to reach support for 3 days"

Format: Markdown table with columns No., Category, Sentiment, Urgency, Justification.

Trend Analysis

Analyze this monthly sales data and identify:

Data:
Jan: 45K | Feb: 42K | Mar: 51K | Apr: 48K | May: 55K | Jun: 62K
Jul: 58K | Aug: 41K | Sep: 67K | Oct: 71K | Nov: 78K | Dec: 85K

1. General trend (growth, decline, stagnation)
2. Anomalies or unusual variations
3. Any seasonality
4. A projection for the next 3 months (with assumptions)

Justify each observation with the data.

Practice

Take a document you need to analyze regularly (report, customer feedback, competitive intelligence). Create a structured summary prompt and a data extraction prompt. Test them on a real document and adjust the output format until you get a result that is directly usable.

Key Takeaways

• Specify the type of summary desired (executive, structured, comparative) and the target audience
• For extraction, explicitly list the fields to extract and the output format
• Request justifications for analyses and classifications
• Specify how to handle missing data (“unspecified” rather than invention)
• A good analysis prompt produces a result that is directly usable without reformatting