# Technical Document Extraction: Data Agent Workflow Comparison
This document provides a comprehensive extraction of the information contained in the provided infographic, which compares a manual data analysis workflow ("Without Data Agent") against an automated AI-driven workflow ("With Data Agent").
## 1. Header Information
* **Query:** "Analyze the **sales trends** of **Arizona** retail data and generate a summary report with **visualizations**."
* *Note: Specific keywords are highlighted in blue in the original image.*
---
## 2. Workflow Comparison: "Without Data Agent"
This section describes the traditional, manual process of data analysis.
### A. Input Data Table
| Date | Region | Product | Sales ($) |
| :--- | :--- | :--- | :--- |
| 2021-01-15 | Arizona | Electronics | 9,230 |
| 2021-01-18 | Arizona | Clothing | 5,600 |
| ... | ... | ... | ... |
| 2021-03-20 | Utah | Grocery | 8,200 |
### B. Process Steps
1. **Manual Designed SQL:** A user manually writes a query.
* *Text:* "Select ... Where Region = 'Arizona' Group by ... Order By ..."
2. **Data Processing & Visualization:** Manual coding or tool usage.
* *Text:* "(Pandas / Excel / Matplotlib)"
3. **Output:** A line chart titled "Arizona Retail Sales Trends."
---
## 3. Workflow Comparison: "With Data Agent"
This section details the automated pipeline using an AI agent.
### A. Initial Processing (Left Column)
* **Perception (Query & Data Understanding):**
* **Intention:** Analyze the sales trends
* **Region:** Arizona
* **Output:** visualized report
* **Planning + Decomposition:** The agent breaks the query into a list of **Subtasks**:
1. Identify Data Source
2. Generate SQL
3. Execute SQL
4. Verify Results
5. Visualization
6. Summarize Findings
### B. Execution Pipeline (Right Column)
The agent executes the subtasks sequentially:
1. **Identify Data Source:** References the same input table as the manual process.
2. **Generate SQL:**
* *Code:* `SELECT DATE_TRUNC('month', Date) AS month, SUM(Sales) FROM retail_sales WHERE Region = 'Arizona'`
3. **Execute SQL:** Filters the data specifically for Arizona.
* *Result Table Snippet:* Includes 2021-01-15 (Electronics), 2021-01-18 (Clothing), and 2021-03-06 (Arizona, Food, 8,900).
4. **Verify Results:** Uses an LLM (represented by the OpenAI logo) to check accuracy.
* *Options:* "Need Refinement" (loops back to step 1) or "Correct" (proceeds to step 5).
5. **Visualization:** Generates the final chart.
6. **Summary:** Generates a natural language report.
* *Text:* "Arizona retail sales decreased in Feb 2021 but rebounded strongly in March, ending the quarter with a 12% growth compared to January."
---
## 4. Data Visualization Analysis
Both workflows produce a line chart titled **"Arizona Retail Sales Trends"**.
### Chart Components
* **Y-Axis:** Sales (K). Scale: 0 to 30.
* **X-Axis:** Time. Scale: 1980 to 2020 (Note: This appears to be placeholder/template data as it contradicts the 2021 query dates).
* **Legend/Labels:** Placed at the right end of the lines [x=far right, y=aligned with line ends].
* **Data Series & Trends:**
1. **Electronic (Orange Line):** Starts at ~15K, shows significant volatility with a sharp upward trend, ending as the highest category at ~28K.
2. **Clothing (Light Blue Line):** Relatively flat trend with minor fluctuations, hovering around the 12K-15K mark.
3. **Grocery (Medium Blue Line):** Flat trend, consistently positioned around 8K-10K.
4. **Food (Dark Blue Line):** The lowest value series, remaining flat near the 5K mark.
---
## 5. Component Isolation Summary
* **Header:** Defines the natural language query.
* **Top Row:** Shows the "Manual" path: Data -> SQL -> Code -> Chart.
* **Bottom Left:** Shows the "Agent Logic": Perception -> Planning -> Subtask list.
* **Bottom Right:** Shows the "Agent Execution": Step-by-step data filtering, SQL generation, LLM verification, and final summarized output.
