Classification Workflows

The NAICS MCP Server supports several classification workflows depending on your needs.

Simple Classification

Use case: Quick classification of a single business

User: "What NAICS code for a pizza restaurant?"

AI: Uses classify_business → Returns 722511 (Full-Service Restaurants)

Steps:

1. Call classify_business with the description
2. Return the recommended code with confidence
3. Note any cross-reference warnings

Exploratory Classification

Use case: Understanding options when unsure

User: "I need to classify a company that sells software online"

AI: Uses search_naics_codes → Returns multiple options
    Uses get_code_hierarchy → Explains each option
    Asks clarifying questions

Steps:

1. Call search_naics_codes with limit=5
2. For top results, call get_code_hierarchy
3. Present options with context
4. Ask clarifying questions if needed
5. Let user select or provide more info

Validation Workflow

Use case: Verifying an existing classification

User: "Is 541511 correct for my web development agency?"

AI: Uses validate_classification → Checks fit
    Uses get_cross_references → Checks exclusions

Steps:

1. Call validate_classification with description and code
2. If valid, confirm with any caveats
3. If not valid, suggest alternatives
4. Check cross-references for edge cases

Batch Classification

Use case: Processing multiple businesses

User: "Classify these 50 businesses from my spreadsheet"

AI: Uses classify_batch → Returns all classifications
    Flags low-confidence items for review

Steps:

1. Call classify_batch with all descriptions
2. Group results by confidence level
3. Present high-confidence results
4. Flag low-confidence items for user review

Audit Trail Workflow

Use case: Compliance documentation

User: "Classify this business and save it for our records"

AI: Uses classify_business → Gets classification
    Uses write_to_workbook → Saves decision

Steps:

1. Call classify_business with full analysis
2. Present recommendation to user
3. Upon confirmation, call write_to_workbook
4. Include reasoning and alternatives considered

Hierarchy Exploration

Use case: Understanding NAICS structure

User: "Show me all subcategories under Retail Trade"

AI: Uses get_children repeatedly → Builds tree

Steps:

1. Start with sector code (e.g., “44-45”)
2. Call get_children to get subsectors
3. Optionally drill down further
4. Present as structured hierarchy

Comparison Workflow

Use case: Deciding between similar codes

User: "What's the difference between 722511 and 722513?"

AI: Uses compare_codes → Shows side-by-side
    Uses get_cross_references → Shows boundaries

Steps:

1. Call compare_codes with both codes
2. Highlight key differences
3. Call get_cross_references for each
4. Explain when to use each code

Cross-Sector Analysis

Use case: Discovering when a business might belong in a different sector

User: "Classify this furniture wholesaler"

AI: Uses classify_business → Gets initial code
    Uses get_relationship_stats → Checks for cross-sector alternatives
    Uses get_cross_sector_alternatives → Shows options in other sectors
    Compares retail vs wholesale codes

Steps:

1. Call classify_business with description
2. Call get_relationship_stats on the top result
3. If has_cross_sector: true:
   • Call get_cross_sector_alternatives to see options
   • Call compare_codes for top alternatives
   • Call get_cross_references to check exclusions
4. Present alternatives with sector context
5. Let user decide based on PRIMARY activity

Why this matters:

• A “bakery” might be Manufacturing (311), Retail (445), or Food Service (722)
• A “software company” might be Information (51) or Professional Services (54)
• Cross-sector alternatives surface these boundary cases for human evaluation

Interpreting similarity scores: