AI Classification

What Is AI Classification?

AI Classification is Go Fig’s machine learning capability that automatically categorizes and tags your data. Instead of manually coding every transaction, expense, or record, AI Classification learns your categorization patterns and applies them consistently at scale.

Common applications:

• Expense categorization (travel, supplies, software, etc.)
• Revenue classification by type or product
• Customer segmentation
• Vendor categorization
• Transaction tagging
• Document classification

How Does AI Classification Work?

Learning Phase

AI Classification learns from your existing data:

1. Analyze historical data: Reviews how transactions were previously categorized
2. Identify patterns: Finds correlations between transaction attributes and categories
3. Build models: Creates classification rules based on patterns
4. Validate accuracy: Tests predictions against known categorizations

Classification Phase

Once trained, the system classifies new data:

1. Ingest new records: New transactions or data enters Go Fig
2. Extract features: Analyzes relevant attributes (vendor name, amount, description, etc.)
3. Predict category: Applies learned patterns to suggest classification
4. Confidence scoring: Indicates how certain the prediction is
5. Human review: Low-confidence items flagged for manual review

Continuous Improvement

The system gets smarter over time:

• Learns from corrections you make
• Adapts to new vendors, products, and patterns
• Alerts you to classification drift
• Suggests rule refinements

AI Classification Use Cases

Expense Categorization

Problem: Thousands of credit card transactions need GL coding every month.

Solution: AI Classification:

• Learns from historical expense coding
• Automatically categorizes 85%+ of transactions
• Flags unusual items for review
• Ensures consistent treatment of similar expenses

Result: Hours of manual coding reduced to minutes of exception review.

Revenue Recognition

Problem: Revenue needs classification by type for proper recognition treatment.

Solution: AI Classification:

• Analyzes order attributes (product, customer, terms)
• Applies appropriate revenue category
• Flags complex transactions for review
• Maintains audit trail of classification logic

Result: Consistent revenue classification with documented rationale.

Vendor Categorization

Problem: New vendors constantly added; categorization inconsistent.

Solution: AI Classification:

• Learns vendor categorization patterns
• Suggests categories for new vendors
• Identifies miscategorized existing vendors
• Maintains vendor master data quality

Result: Clean vendor data supporting better spend analysis.

Customer Segmentation

Problem: Customers need segmentation for reporting and analysis.

Solution: AI Classification:

• Analyzes customer attributes and behavior
• Assigns segment labels automatically
• Updates segments as customer profiles change
• Enables segment-based reporting

Result: Dynamic customer segmentation without manual maintenance.

Confidence Levels and Human Review

AI Classification provides confidence scores with every prediction:

High Confidence (>90%)

• Automatically applied
• Included in standard reporting
• Logged for audit purposes

Medium Confidence (70-90%)

• Applied with flag for review
• Included in reports with notation
• Queued for periodic validation

Low Confidence (<70%)

• Not automatically applied
• Routed to human reviewer
• Used to improve model training

You control the thresholds based on your risk tolerance and audit requirements.

AI Classification vs. Rule-Based Systems

Aspect	Rule-Based	AI Classification
Setup effort	High (define all rules)	Low (learns from data)
New patterns	Requires new rules	Adapts automatically
Edge cases	Often missed	Better handling
Maintenance	Ongoing rule updates	Self-improving
Accuracy	Depends on rule quality	Typically 85-95%
Audit trail	Rule that matched	Factors considered

Data Quality Impact

AI Classification improves overall data quality:

Consistency: Same transactions classified the same way, every time

Completeness: Fewer uncategorized or “miscellaneous” items

Accuracy: Catches human errors and miscategorizations

Timeliness: Classifications applied immediately, not batched monthly

Getting Started

AI Classification is available on Go Fig Growth and Enterprise plans. Implementation involves:

1. Data review: Analyze your historical categorization patterns
2. Model training: Build classification models from your data
3. Validation: Test accuracy against known categorizations
4. Deployment: Enable automatic classification
5. Monitoring: Track accuracy and refine over time

Most customers see 85%+ automatic classification accuracy within the first month.