Data Analyst Agent

What This Agent Does

The Data Analyst Agent transforms raw data files (Excel, CSV, PDF, documents) into rich, interactive dashboards with custom widgets—all through natural language prompts. Upload your data and instantly get graphical analysis with charts and graphs. No coding or complex BI tools required.

Time Savings: Create comprehensive dashboards in 10-15 minutes instead of 2-3 hours in Excel/Tableau/Power BI.

Key Capabilities:

  • ✅ Upload Excel/CSV/PDF and ask natural language questions about your data
  • ✅ Automatic chart and graph generation (bar, line, pie, scatter, heatmaps)
  • ✅ Custom dashboard templates with configurable widgets
  • ✅ ABC branding compliance (colors, fonts, logos automatically applied)
  • ✅ Shareable dashboards via chat links with role-based access
  • ✅ Chat with your dashboards to ask questions about the data behind widgets
  • ✅ Auto-dashboard generation makes insights accessible to everyone

Getting Started

Step 1: Access the Agent

  1. Login to Cortx portal with ABC credentials
  2. Navigate to “Core Agents” → “DataAnalystAgent”
  3. Agent greets you : “Upload your data file and I’ll help you analyze it!

Step 2: Upload Your Data

Supported File Types:

  • Excel: .xlsx, .xlsm, .xls (including multiple sheets)
  • CSV: .csv (comma-separated values)
  • TSV: .tsv (tab-separated values)
  • PDF: Tables extracted from PDF reports
  • Word: Tables from .docx documents
  • Google Sheets: Export and upload

What to Upload:

  • Sales data
  • Financial reports
  • Customer metrics
  • Branch performance data
  • Survey results
  • Transaction logs
  • Any structured data

How to Upload:

Click “Upload File” button → Select file from computer → Wait for confirmation

Agent responds:

“I’ve analyzed your file: Branch_Performance_2025.xlsx

I found:
– 250 records
– 12 columns (Branch ID, Branch Name, Region, Revenue, Costs, Customer Count, NPS Score, etc.)
– Data spans: Jan 2025 – Dec 2025

What would you like to know about this data?”

Step 3: Start Analyzing

You Have Two Options:​

Option A: Ask Agent for Automatic Insights

You: “Give me insights from this data.”

Agent: “Here’s what I found:

Overall Performance:
– Total revenue across all branches: $47.2M
– Average branch revenue: $188.8K
– Top performing region: North (avg $215K per branch)
– Lowest performing region: South (avg $162K per branch)

Key Insights:
– 15 branches (60%) exceeded revenue targets
– 8 branches (32%) are underperforming (below 85% of target)
– Strong correlation between customer count and revenue (r=0.87)
– NPS scores range from 42 to 68 (average: 55)

Areas of Concern:

– 5 branches need urgent attention (revenue < 70% of target)
– South region consistently lower performance across all metrics

Would you like me to visualize any of these insights?”

Agent proactively analyzes and highlights patterns, trends, outliers.

Option B: Ask Specific Questions

You: “Which are the top 5 performing branches by revenue?”

Agent: “Top 5 Branches by Revenue:

1. Downtown Main Branch: $425K (Region: North)
2. Westside Financial Center: $398K (Region: West)
3. Harbor District Branch: $387K (Region: North)
4. Tech Hub Branch: $361K (Region: West)
5. University Avenue Branch: $342K (Region: East)

Would you like to see this as a chart?”

Automatic Chart Generation

Agent automatically creates visualizations:

Example 1: Bar Chart

You: “Show me revenue by region.”

Agent: [Generates bar chart]
– North: $2.8M
– South: $1.6M
– East: $2.1M
– West: $2.4M

[ABC branded colors, proper labels, clean design]

“The North region leads with $2.8M, 17% higher than the overall average.”

Example 2: Line Chart (Trends)

You: “Show me revenue trends over the year.”

Agent: [Generates line chart showing monthly revenue]
– January: $3.2M
– February: $3.4M
– […]
– December: $4.8M

“Revenue shows consistent growth throughout the year with 50% increase from January ($3.2M) to December ($4.8M).
Peak occurred in Q4 (Nov-Dec) coinciding with holiday season.”

Example 3: Pie Chart (Distribution)

You: “What’s the distribution of customer satisfaction ratings?”

Agent: [Generates pie chart] – Excellent (9-10): 35%
– Good (7-8): 42%
– Fair (5-6): 18%
– Poor (1-4): 5%

“77% of customers rate us Good or Excellent. Focus area:

23% in Fair/Poor categories.”

Example 4: Scatter Plot (Correlations)

You: “Is there a relationship between customer count and revenue?”

Agent: [Generates scatter plot with trend line]
– Each point represents one branch
– X-axis: Customer Count
– Y-axis: Revenue
– Trend line shows strong positive correlation (r=0.87)

“Strong correlation: branches with more customers generate proportionally higher revenue. Outliers: 2 high-customer/low-revenue branches may have retention or product penetration issues.”

Example 5: Heatmap

You: “Show me performance across regions and quarters.”

Agent: [Generates heatmap]
            Q1        Q2       Q3      Q4
North $680K $710K $725K $785K [Dark Green]
South $390K $405K $418K $487K [Light Green]
East    $510K $528K $542K $620K [Medium Green]
West   $585K $598K $610K $695K [Dark Green]

“North and West regions consistently outperform. South shows improvement but remains lowest. All regions show Q4 strength.”

Creating Custom Dashboards

Dashboard Templates

Instead of individual charts, create comprehensive dashboards with multiple widgets:

You: “Create a branch performance dashboard.”

Agent: “I’ll create a dashboard with these widgets:
1. Branch Performance Overview (top-level metrics)
2. Top 5 Performing Branches (bar chart)
3. Bottom 5 Branches Needing Attention (bar chart)
4. Regional Comparison (pie chart)
5. Monthly Revenue Trend (line chart)
6. Customer Satisfaction by Branch (scatter plot)

Shall I proceed with this layout?”

You: “Yes, but add a widget showing cost efficiency (revenue vs. costs).”

Agent: [Creates dashboard with 7 widgets, ABC branded]

Widget Types Available

Metric Cards:

  • Single number displays

  • Example: “Total Revenue: $47.2M” with change indicator (+12% vs. last year)

Charts:

  • Bar charts (comparisons)

  • Line charts (trends over time)

  • Pie charts (distributions)

  • Scatter plots (relationships)

  • Heatmaps (multi-dimensional comparisons)

  • Area charts (cumulative trends)

Tables:

  • Sortable data tables

  • Top/bottom N rankings

  • Detailed drill-downs

Text Insights:

  • Written summaries

  • Key findings

  • Recommendations

Dashboard Layout Customization

Define your dashboard structure:

You: “Create a 2-column dashboard:

Left column (wider):
– Large revenue trend line chart (top)
– Regional performance bar chart (bottom)

Right column (narrower):
– Key metrics cards: Total Revenue, Avg Branch Revenue, Top Branch
– Top 5 branches table
– Bottom 5 branches table

Use ABC’s executive dashboard color scheme.”

Agent: [Creates custom layout matching specifications]

ABC-Specific Use Cases

Use Case 1: Branch Performance Analysis

Your Scenario: Monthly review of all retail branch metrics

What You Upload:

Branch_Performance_December_2025.xlsx containing:
– Branch ID and name
– Region
– Revenue (actual vs. target)
– Expenses
– Customer count (active, new, churned)
– NPS scores
– Product penetration rates
– Staff count and productivity metrics

What You Ask:

“Create a comprehensive branch performance dashboard showing:
1. Overall performance summary (total revenue, avg per branch, etc.)
2. Top 10 and bottom 10 branches
3. Regional comparison
4. Revenue vs. target achievement rate by branch
5. Customer satisfaction trends
6. Cost efficiency analysis (revenue per dollar spent)
7. Identify branches needing urgent intervention”

Agent Output:

  • Full dashboard with all requested widgets

  • Automatic insights highlighting concerning trends

  • ABC branded, ready to present to leadership

  • Exportable to PDF or shareable via link

Time Saved: 2-3 hours of Excel work → 10 minutes

Use Case 2: Financial Performance Reporting

Your Scenario: Quarterly financial review for board presentation

What You Upload:

Q4_2025_Financial_Results.xlsx with:
– Revenue by product line
– Expenses by category
– Profit margins
– Year-over-year comparisons
– Budget vs. actual
– Geographic breakdown

What You Ask:

“Create a financial performance dashboard for Q4 2025 board meeting:
– Executive summary (revenue, profit, margins)
– Revenue breakdown by product line (show top 5)
– Expense analysis (variance from budget)
– Profitability by region
– Quarterly trend (Q1-Q4 comparison)
– Year-over-year growth metrics

Highlight any areas >10% variance from budget or forecast.”

Agent Output:

  • Board-ready financial dashboard

  • Variances highlighted in red/green

  • Explanatory text for key metrics

  • One-click export to PDF for pre-read materials

Use Case 3: Customer Behavior Analysis

Your Scenario: Understanding customer segmentation and behavior patterns

What You Upload:

Customer_Data_2025.csv with:
– Customer demographics (age, income, location)
– Account types and balances
– Transaction history (frequency, amounts, types)
– Digital engagement (app usage, web logins)
– Product holdings (checking, savings, loans, credit cards)
– Lifetime value calculations

What You Ask:

“Analyze customer segments and create dashboard showing:
1. Customer distribution by value tier (high/medium/low value)
2. Product penetration by segment
3. Digital vs. branch usage patterns
4. Churn risk indicators
5. Cross-sell opportunities
6. Customer lifetime value distribution

Focus on actionable insights for product marketing team.”

Agent Output:

  • Segmentation dashboard with clear customer profiles

  • Opportunity sizing (e.g., “40% of high-value customers have only 1 product – cross-sell opportunity of $12M annual revenue”)

  • Behavioral patterns visualized

  • Churn risk scoring with early warning indicators

Use Case 4: Campaign Performance Tracking

Your Scenario: Marketing campaign results analysis

What You Upload:

Campaign_Results_Q4.xlsx:
– Campaign name and type
– Channels used (email, social, digital ads, branch)
– Budget spent
– Impressions/reach
– Clicks/engagement
– Leads generated
– Conversions
– Revenue generated
– Customer acquisition cost
– ROI by campaign

What You Ask:

“Create a campaign performance dashboard showing:
– Overall campaign ROI and performance
– Best performing campaigns (by ROI, conversions, engagement)
– Channel effectiveness comparison
– Cost per acquisition by campaign
– Budget efficiency (actual spend vs. allocated)
– Recommendations for Q1 budget allocation based on Q4 performance

Rank campaigns by ROI and flag any with negative returns.”

Agent Output:

  • Campaign scorecard

  • Channel mix recommendations

  • Budget reallocation suggestions based on performance

  • Clear winners and losers identified

Use Case 5: Operational Efficiency Metrics

Your Scenario: Process improvement analysis across operations

What You Upload:

Operations_Metrics_2025.xlsx:
– Process completion times
– Error rates
– Staff productivity metrics
– Queue/wait times
– SLA compliance rates
– Resource utilization
– Cost per transaction

What You Ask:

“Analyze operational efficiency and create dashboard showing:
1. Process time averages and trends
2. SLA compliance by process type
3. Error rate analysis (identify high-error processes)
4. Staff productivity comparison
5. Cost efficiency metrics
6. Bottleneck identification
7. Improvement opportunity sizing

Focus on processes where we’re underperforming vs. benchmarks.”

Agent Output:

  • Operations health dashboard

  • Bottlenecks clearly identified

  • Improvement opportunities quantified (e.g., “Reducing loan processing time by 1 day saves $450K annually”)

  • Benchmarking against industry standards

Interactive Features

Chat with Your Dashboard

Once dashboard is created, you can ask questions about the data:

Dashboard Created: “Branch Performance Dashboard”

You: “Which branch has the highest customer satisfaction?”

Agent: “University Avenue Branch has the highest NPS score of 68. They serve 450 customers with average account balance of $12,400. This branch consistently ranks in top 3 for satisfaction.”

You: “What’s driving their high satisfaction?”

Agent: “Analyzing branch attributes:
– Longest average tenure of staff (5.2 years vs. 3.1 avg)
– Highest product cross-sell rate (3.2 products per customer vs. 2.1 avg)
– Shortest average wait time (4.3 minutes vs. 8.7 avg)
– Above-average digital adoption (72% vs. 58% avg)

Key driver appears to be experienced staff + efficient operations.”

Continuous Conversation:

  • Ask follow-up questions

  • Drill into specific data points

  • Request different visualizations

  • Export specific findings

Suggested Questions

Agent proactively suggests relevant questions you might want to ask:

Agent: “Here are some questions you might want to explore:

Performance Analysis:

– Which branches improved the most compared to last quarter?
– What’s the correlation between staff count and revenue?
– Show me branches underperforming by more than 20%

Financial Insights:
– What’s the cost-to-income ratio by region?
– Which branches have the best profit margins?
– Show revenue per employee analysis

Customer Insights:
– What’s the relationship between NPS scores and revenue?
– Which branches are losing customers?
– Customer growth rate by region

Click any question or ask your own!”

Sharing Dashboards

Share with Team

You: “Share this dashboard with the regional managers team.”

Agent: “Who should have access?
– View Only (can see dashboard, can’t modify)
– Comment (can add notes and questions)
– Edit (can modify widgets and layout)”

You: “Give view access to all regional managers.
Give edit access to my direct reports.”

Agent: “Dashboard shared!
– View access: Regional Managers group (23 people)
– Edit access: Sarah Johnson, Mike Chen, Priya Sharma

Share link: https://cortx.abc.com/dashboard/branch-perf-q4-2025

Team members will receive email notification.”

Collaboration Features

Comments on Widgets:

Viewer adds comment on “Bottom 5 Branches” widget:
“South Bay branch needs immediate attention – revenue down 30% vs. last quarter. Schedule intervention meeting for next week.”

Dashboard owner receives notification and can respond.

Version History:

Dashboard: Branch Performance Q4 2025
– v1.0 (Jan 15): Initial creation
– v1.1 (Jan 16): Added cost efficiency widget
– v1.2 (Jan 18): Updated with latest data, added regional breakdown
– v1.3 (Jan 20): Added year-over-year comparison [CURRENT]

Scheduled Updates:

You: “Update this dashboard automatically every Monday morning with the latest data from the shared folder.”

Agent: “Scheduled! Dashboard will refresh every Monday at 8am using data from /shared/branch-performance/weekly-reports/”

Advanced Features

Multiple Data Sources

Combine data from multiple files:

You upload:
– Branch_Revenue.xlsx
– Branch_Costs.xlsx
– Customer_Satisfaction.csv
– Staff_Metrics.xlsx

You: “Merge these files using Branch ID as the key.
Create a comprehensive dashboard combining all metrics.”

Agent: [Merges data, creates unified dashboard]
“I’ve combined all data sources. 250 branches matched across all files. Dashboard now shows revenue, costs, profit, customer metrics, and staff productivity in one view.”

Calculated Metrics

Create new metrics from existing data:

You: “Calculate and add these metrics:
– Profit Margin = (Revenue – Costs) / Revenue * 100
– Revenue per Employee = Revenue / Staff Count
– Customer Acquisition Cost = Marketing Spend / New Customers
– Customer Lifetime Value = Avg Revenue per Customer * Avg Tenure

Add these as new columns and create visualizations.”

Agent: [Calculates metrics, adds to dashboard]

Filtering and Drill-Down

Add interactive filters:

You: “Add filters to this dashboard:
– Region selector (North, South, East, West, All)
– Time period selector (Month, Quarter, Year)
– Performance tier (High, Medium, Low performing branches)

When user changes filter, update all widgets automatically.”

Agent: [Adds interactive filters]

User Experience:
User selects “North Region” + “Q4” + “Low Performing”
→ All charts update to show only North region low performers in Q4
→ Insights regenerate for filtered view

Benchmarking

Compare against targets or benchmarks:

You: “Add benchmark comparison:
– Target revenue per branch: $200K
– Target NPS score: 60
– Industry average profit margin: 18%

Show each branch/metric vs. benchmark.
Use green for above target, red for below.”

Agent: [Adds benchmark lines/indicators to all charts]
“15 branches above revenue target (green),
10 branches below (red), shown with benchmark line at $200K.”

Export Options

Multiple ways to share insights:

PDF Export:

“Export this dashboard as PDF for executive team review.”
→ High-quality PDF with all charts and data

Excel Export:

“Export the underlying data to Excel with all calculated metrics.”
→ Excel file with all data, calculations, pivot tables

PowerPoint Export:

“Create a PowerPoint presentation with key charts from this dashboard.”
→ PPT with one chart per slide, ABC branded

Image Export:

“Export the regional performance chart as PNG for the newsletter.”
→ High-res image file

ABC Branding

DO: Provide Rich Input Context

All dashboards automatically apply ABC brand standards:

✅ Color Palette:

  • Primary colors for main data series

  • Accent colors for highlights

  • Consistent color scheme across all charts

✅ Typography:

  • ABC fonts for titles, labels, legends
  • Proper hierarchy and sizing

✅ Layout:

  • Clean, professional spacing

  • ABC logo on dashboards

  • Consistent widget styling

✅ Chart Styling:

  • Professional gridlines

  • Clear axis labels

  • Readable legends

  • Appropriate chart types for data

You Don’t Need to Think About Design: Agent handles all formatting automatically. Every dashboard is presentation-ready.

Best Practices

DO: Clean Your Data Before Upload

✅ Good Data:

  • Consistent column headers

  • No merged cells

  • Dates in standard format

  • No blank rows in middle of data

  • Clear column names (not “Column1”, “Column2”)

❌ Problematic Data:

  • Mixed formats (some dates as text, some as numbers)

  • Subtotals mixed with data rows

  • Multiple tables on one sheet

  • Inconsistent naming (North, north, N, NORTH)

Quick Cleanup:

If your data has issues:

1. Remove subtotal rows
2. Standardize column names
3. Convert dates to proper format 4. Remove merged cells
5. Delete empty rows

Or ask agent: “Can you clean this data first?”
Agent will attempt basic cleaning.

DO: Start with Simple Questions

Progressive Approach:

Step 1: “Show me total revenue by region.”
→ Understand data structure

Step 2: “Now break down North region by individual branches.”
→ Drill into details

Step 3: “For top 5 North branches, show monthly trend.”
→ Deep analysis

Step 4: “Create comprehensive North region dashboard.”
→ Full insights

❌ Don’t Jump to Complex Immediately:

“Create a multi-dimensional analysis showing revenue, costs, profit, customer metrics, staff productivity, year-over-year growth,
regional benchmarking, and predictive forecasting all in one dashboard.”

→ Too complex to start. Build incrementally.

DO: Provide Context

✅ Good:

“Analyze this sales data. We’re particularly interested in:
– Q4 performance (holiday season impact)
– Product category trends (electronics vs. home goods)
– Regional differences (urban vs. suburban branches)
– Effect of new loyalty program launched in October

Flag any unusual patterns.”

❌ Too Generic:

“Analyze this file.”
→ Agent doesn’t know what you care about

DO: Iterate on Dashboards

Workflow:

Version 1: “Create basic revenue dashboard”
→ Agent creates initial version

Version 2: “Add customer satisfaction metrics”
→ Agent adds new widgets

Version 3: “Replace pie chart with stacked bar chart”
→ Agent modifies visualization

Version 4: “Add year-over-year comparison”
→ Agent enhances with historical context

Final: “Perfect! Share with team.”

Don’t expect perfection on first try. Build and refine.

DON'T: Overload One Dashboard

⚠️ Too Many Widgets:

Dashboard with 25 widgets showing every possible metric
→ Overwhelming, hard to find insights, slow loading

✅ Better Approach:

Create multiple focused dashboards:
– Executive Summary Dashboard (5-6 key widgets)
– Regional Performance Dashboard (focus on regions)
– Customer Insights Dashboard (customer metrics)
– Operational Efficiency Dashboard (process metrics)

Each serves a specific purpose and audience.

DON'T: Ignore Data Quality Warnings

Agent: ” ⚠️ Warning: I found 15 branches with missing cost data.
These branches will show incomplete profit calculations.
Would you like to:
1. Exclude them from analysis
2. Mark them as ‘incomplete’ in charts
3. Estimate costs based on similar branches”

DON’T: Ignore and proceed

DO: Address data quality issue before analyzing

Prompting Examples

Quick Analysis

“What are the key insights from this data?”
“Show me the top 10 [metric] by [dimension]”
“What trends do you see in [specific column]?”
“Identify any outliers or unusual patterns”
“Compare [metric A] vs [metric B]”

Visualization Requests

“Create a bar chart showing [metric] by [dimension]”
“Show me [metric] trend over time as a line chart”
“Create a pie chart of [dimension] distribution”
“Plot the relationship between [X] and [Y] as a scatter plot”
“Make a heatmap showing [rows] vs [columns]”

Dashboard Creation

“Create a [topic] dashboard with:
– [Widget 1 description]
– [Widget 2 description]
– [Widget 3 description]

Target audience: [who will use this]
Focus: [what they care about most]”

Comparative Analysis

“Compare [Group A] vs [Group B] across all metrics”
“Show me how [current period] compares to [past period]”
“Benchmark our performance against [target/industry standard]”
“Rank all [entities] by [metric] and show top/bottom 10”

Filtering and Segmentation

“Show me only [specific segment] data”
“Filter to [condition], then analyze”
“Break down [metric] by [dimension 1] and [dimension 2]”
“Exclude [specific values] from analysis”

Troubleshooting

Issue: "Agent can't read my Excel file"

Common Causes:

  • Password-protected file

  • Corrupted file

  • Very old Excel format (.xls from Excel 2003)

  • File too large (>50MB)

Solutions:

1. Remove password protection
2. Save as .xlsx (modern format)
3. For large files: Filter to recent data only, or split into smaller files
4. Check file isn’t corrupted (can you open it in Excel?)

Issue: "Charts look wrong / data doesn't make sense"

Common Causes:

  • Data interpreted incorrectly (numbers stored as text)

  • Wrong aggregation (summing when should average)

  • Duplicates in data

  • Mismatched joins between files

Solutions:

“I think you’re summing customer counts when you should average them.
Please recalculate using average instead.”

Or: “Check for duplicate Branch IDs in the data.
I think some branches are counted twice.”

Or: “Revenue column is stored as text. Convert to numbers first.”

Issue: "Dashboard is too slow to load"

Causes:

  • Too much data (millions of rows)

  • Too many widgets (15+)

  • Complex calculations

Solutions:

Option 1: “Aggregate data to monthly level instead of daily.”
Option 2: “Filter to most recent 6 months only.”
Option 3: “Split into multiple dashboards instead of one huge dashboard.” Option 4: “Remove widgets that are rarely viewed.”

Issue: "Dashboard doesn't update with new data"

Solutions:

Manual: “Refresh this dashboard with the latest data file.”
[Upload updated file]

Automatic: “Set up automatic refresh from this folder.”
Agent will pull latest file from designated location.

Integration with Other Agents

With PPT Agent - Present Insights

Workflow:
1. Create dashboard in Data Analyst Agent
2. Export key charts as images
3. Open PPT Agent:

“Create a presentation showing our Q4 branch performance analysis.

Use these charts from my dashboard:
[upload chart images]

Create slides covering:
– Executive summary (key metrics)
– Regional performance comparison
– Top performers analysis
– Areas needing attention
– Recommendations

Audience: Leadership team
10 slides maximum”

With Document Agent - Create Written Report

Workflow:
1. Generate insights in Data Analyst Agent
2. Open Document Agent:

“Create a written analysis report from this dashboard data.

[Share dashboard link or export Excel]

Include:
– Executive summary
– Key findings
– Detailed regional analysis
– Trend analysis
– Recommendations
– Appendix with data tables

Format: ABC Standard Analysis Report Template”

With Vendor Evaluation Agent - Analyze Vendor Data

Workflow:
1. Create vendor comparison dashboard
2. Use insights to inform evaluation:

“Here’s performance data on our current vendors.
[Share dashboard]
Use this to inform the vendor evaluation analysis.”

Getting Help

In-Platform:

  • Help icon (bottom right)

  • In-app AI assistant

  • Sample datasets to practice with

Resources:

Common Questions:

  • “What file size limits exist?” → 50MB per file recommended

  • “Can I connect to live databases?” → Yes, via custom integrations (contact IT)

  • “How often can dashboards refresh?” → Real-time, hourly, daily, weekly, or manual

End of Document Generation Agent User Guide

 

Was this article helpful?

On this page