analyze-report

name: analyze-report description: "Analyze Snowflake AI cost data and generate insights and reports." parent_skill: ai-cost-dashboard

Analyze / Report AI Costs

When to Load

From ai-cost-dashboard SKILL.md when user selects Analyze / Report intent.

Prerequisites

• AI Cost Dashboard is deployed (tables exist in ADMIN_DB.AI_COSTS)
• Snowflake connection is active

Data Sources

Service categories in unified table: LLM Functions, Analyst, Document Processing, Fine Tuning, Provisioned Throughput, REST API, Search, Document AI, LLM Functions Query, Cortex Code CLI

Workflow

Step 1: Determine Analysis Scope

Goal: Understand what the user wants to analyze.

Actions:

1. Ask user for analysis parameters:

   What would you like to analyze?
   
   1. Overall AI credit consumption
   2. Cost breakdown by service category
   3. Cost trends and anomalies
   4. Cost attribution by user
   5. Model-level token usage
   
   Time range: Last 7 / 30 / 90 / 365 days / Custom
   

2. Optionally refresh data first:

   CALL ADMIN_DB.AI_COSTS.REFRESH_CORTEX_AI_COSTS(<days>);
   CALL ADMIN_DB.AI_COSTS.REFRESH_USER_AI_COSTS(<days>);
   

Output: Defined analysis scope and time range

Step 2: Execute Cost Queries

Goal: Retrieve AI cost data.

Actions:

1. Overall consumption:

   SELECT SERVICE_CATEGORY, SUM(CREDITS) AS TOTAL_CREDITS, SUM(TOKENS) AS TOTAL_TOKENS, SUM(REQUEST_COUNT) AS TOTAL_REQUESTS
   FROM ADMIN_DB.AI_COSTS.CORTEX_AI_UNIFIED_COSTS
   WHERE USAGE_DATE >= DATEADD('day', -<days>, CURRENT_DATE())
   GROUP BY SERVICE_CATEGORY
   ORDER BY TOTAL_CREDITS DESC;
   

2. Daily trend:

   SELECT USAGE_DATE, SUM(CREDITS) AS DAILY_CREDITS
   FROM ADMIN_DB.AI_COSTS.CORTEX_AI_UNIFIED_COSTS
   WHERE USAGE_DATE >= DATEADD('day', -<days>, CURRENT_DATE())
   GROUP BY USAGE_DATE ORDER BY USAGE_DATE;
   

3. Top users:

   SELECT FULL_NAME, SUM(CREDITS) AS TOTAL_CREDITS, COUNT(DISTINCT QUERY_ID) AS REQUESTS
   FROM ADMIN_DB.AI_COSTS.USER_AI_COSTS
   WHERE USAGE_DATE >= DATEADD('day', -<days>, CURRENT_DATE())
   GROUP BY FULL_NAME ORDER BY TOTAL_CREDITS DESC LIMIT 20;
   

4. Model usage:

   SELECT MODEL_NAME, SUM(TOKENS) AS TOTAL_TOKENS, SUM(CREDITS) AS TOTAL_CREDITS
   FROM ADMIN_DB.AI_COSTS.CORTEX_AI_UNIFIED_COSTS
   WHERE USAGE_DATE >= DATEADD('day', -<days>, CURRENT_DATE())
   GROUP BY MODEL_NAME ORDER BY TOTAL_TOKENS DESC;
   

Output: Raw cost data

Step 3: Generate Insights

Goal: Analyze data and surface key findings.

Actions:

1. Calculate key metrics:

   • Total credits consumed in period
   • Daily average credit consumption
   • Top cost-driving services and models
   • Most active users
   • Trend direction (increasing/decreasing/stable)

2. Identify anomalies:

   • Daily spikes above 2x average
   • Unusual service patterns
   • New cost sources appearing

3. Compile findings

Output: Analysis summary

Step 4: Present Report

Goal: Deliver findings to user.

Actions:

1. Present structured report:

   Cortex AI Cost Analysis Report
   Period: <start> to <end>
   
   Summary:
   - Total credits: X
   - Total tokens: X
   - Total requests: X
   - Daily average: X credits/day
   
   Top Services:
   1. <category> - X credits (Y%)
   2. <category> - X credits (Y%)
   
   Top Users:
   1. <name> - X credits, N requests
   
   Top Models:
   1. <model> - X tokens
   
   Anomalies:
   - [Any spikes or unusual patterns]
   
   Recommendations:
   - [Actionable suggestions]
   

2. STOP: Ask if user wants deeper analysis on any area.

3. If requested, drill down and repeat from Step 2 with narrower scope.

Output: Delivered report

Stopping Points

• After Step 1: Confirm analysis scope
• After Step 4: Before any follow-up deep dives

Output

AI cost analysis report with insights, trends, anomalies, and recommendations.