Light's AI automatically categorizes expenses and assigns them to GL accounts based on receipt content and merchant information. Machine learning improves accuracy over time as the system learns your spending patterns.
#How AI Categorization Works
When you submit a receipt, Light's AI:
- Analyzes Receipt Content: Reads merchant name, items, and amount
- Identifies Merchant: Matches to known merchants (airlines, hotels, restaurants)
- Predicts Category: Based on merchant type and items
- Suggests GL Account: Maps category to appropriate GL account
- Assesses Confidence: Shows how certain the AI is about the categorization
- Learns from Corrections: Improves when you correct predictions
Good to know: Light's AI achieves 92%+ accuracy on standard receipts, with accuracy improving as you submit more expenses.
#Spending Categories
Light includes standard categories:
Travel
- Flights
- Hotels
- Ground transportation (taxi, Uber, parking)
- Rental cars
- Train/bus travel
Meals
- Restaurant meals
- Coffee/beverages
- Work meals/working lunches
- Client entertainment
Office Supplies
- Paper, pens, office equipment
- Computer accessories
- Furniture and fixtures
- Technology equipment
Professional Development
- Training courses
- Conferences
- Books and subscriptions
- Certifications
Client Entertainment
- Client meals
- Gifts under threshold
- Event attendance for clients
Other
- Miscellaneous expenses
- Unusual categories not fitting above
Your company can create custom categories tailored to your business.
#AI Category Suggestions
#Review Suggested Category
After submitting receipt:
-
Light shows Category Suggestion:
- Suggested Category: AI's recommendation (e.g., "Travel - Hotel")
- Confidence Level: High/Medium/Low certainty
- Reason: Why Light thinks this category
-
Review the suggestion:
- Looks right: Click Confirm to accept
- Wrong category: Click Change to select correct one
#Overriding Suggestions
If Light categorizes incorrectly:
- Tap Change Category or Edit
- Browse or search available categories
- Select correct category
- Light acknowledges and learns from correction
- Future similar expenses improve
#GL Account Mapping
Each spending category maps to GL accounts:
Travel → GL Accounts
- 6100 Travel Expenses
- 6110 Airfare
- 6120 Hotels
- 6130 Ground Transportation
Meals → GL Accounts
- 6200 Meals & Entertainment
- 6210 Meals & Incidentals
- 6220 Client Entertainment
When you categorize expense:
- Light automatically assigns GL account based on category
- GL account shows on expense record
- Finance team can see which account expense will post to
- You can override GL account if needed
#Merchant Recognition
Light maintains database of common merchants:
Recognized Merchants (no category needed):
- Delta Air Lines → Travel - Airfare
- Hilton Hotels → Travel - Hotel
- Uber → Travel - Ground Transportation
- Starbucks → Meals
- Apple → Technology (or varies by context)
Unrecognized Merchants:
- Small local businesses
- One-time vendors
- International merchants not in database
Light asks you for category for unrecognized merchants.
#Training the AI
Light improves through your feedback:
#Corrections Help Learning
When you correct a category:
- Light notes: "User corrected Starbucks from Meals to Office Supplies"
- Context: Receipt showed office supplies, not food
- AI learns: In future, when Starbucks receipt shows office supplies, categorize differently
- Database updates: Helps other employees with similar expenses
#Frequency Improves Accuracy
The more you use Light:
- After 10-20 expenses: AI understands your category preferences
- After 50+ expenses: Accuracy approaches 95%+
- System adapts: Learns what you categorize as what
- Personalization: Expenses in your context are categorized better
#Company-Level Learning
Across your company:
- Light learns from all employee corrections
- All employees benefit from collective learning
- Common mistakes are caught early
- Shared merchants improve across company
#Custom Categories
#Creating Custom Categories
For company-specific needs:
-
Navigate to Settings > Spending Categories
-
Click Create Category
-
Enter:
- Category Name: (e.g., "Contractor Services")
- Description: What's included
- GL Account: Where expenses post
- Spending Limit: Max per expense (if applicable)
- Requires Approval: If over threshold
-
Click Create
-
AI learns the new category over time
#Custom Category Examples
- Software Subscriptions: Monthly SaaS tools
- Contractor Payments: 1099 contractors and freelancers
- Client Reimbursements: Amounts to reimburse clients
- Equipment Maintenance: Repairs and service contracts
Add categories that match your business.
#Confidence Scores
Light shows how confident the AI is:
- High (Green): 90%+ confident in category
- Medium (Yellow): 70-89% confident
- Low (Red): Below 70% confident
#Acting on Confidence Levels
High Confidence:
- Trust the suggestion
- Auto-approve if policy allows
- Rarely needs correction
Medium Confidence:
- Review the suggestion
- Correct if needed
- Provides feedback
Low Confidence:
- Manual category selection required
- Light can't confidently categorize
- You must provide the category
#Policy-Based Categorization
Enforce company rules through categories:
#Spending Limits by Category
Each category can have limits:
- Travel - Meals: $75 per meal
- Client Entertainment: $500 per event
- Office Supplies: $100 per receipt
Light flags expenses exceeding limits for manager review.
#Pre-Approval Rules
Some categories require pre-approval:
- Professional Development: Over $500
- Client Entertainment: Over $250
- Contractor Services: Any amount
Light routes these to manager automatically.
#Merchant-Level Overrides
Set category preferences by merchant:
- Navigate to Settings > Merchant Rules
- Add merchant name (e.g., "Starbucks")
- Set default category: What to use for this merchant
- Set spending limit: Cap per transaction
- Light applies rule to all Starbucks receipts from all employees
Useful for common merchants with consistent spending patterns.
#Categorization Analytics
Track categorization patterns:
-
Navigate to Reports > Spending by Category
-
View:
- Total spending by category
- Trend: Is category increasing/decreasing
- Top merchants: Who employees buy from most
- Accuracy: How often AI categorization was correct
-
Filter by:
- Employee
- Department
- Time period
- Category
Use to understand spending patterns and improve policies.
#Improving Accuracy
Tips for better AI categorization:
#Providing Clear Receipts
- Good receipt photos: Merchant name and items clear
- Details visible: Itemized breakdown (not just total)
- Non-blurry: Clear, legible text
Light extracts better information from clear receipts.
#Adding Descriptions
-
When submitting, include brief note:
- "Client dinner - Acme Corp"
- "Conference registration"
- "Travel software subscription"
-
AI uses description context
-
Categorization improves
#Correcting Consistently
- If Light suggests wrong category, correct it
- Be consistent in your corrections
- System learns your preferences
- Future suggestions improve
#Handling Edge Cases
Some expenses don't fit standard categories:
Multi-Category Expenses:
- Lunch with supplies shopping (Meals + Office Supplies)
- Light lets you split across categories
- Each portion can have different GL account
One-Time Unusual Expenses:
- Once-off payment not fitting standard categories
- Choose closest category or create custom
- Light notes it as unusual for manager review
#Related Articles