PAINTLOGIC AI-POWERED FEATURES DISCLAIMER

Version: 1.0
Effective Date: [INSERT DATE]
Last Updated: [INSERT DATE]

Plain English Summary

PaintLogic uses artificial intelligence to help you work faster—parsing receipts, matching projects, and extracting data. These AI features are tools to assist your professional judgment, not replace it. Always verify AI outputs before making business decisions.

1. AI Technology Disclosure

PaintLogic uses artificial intelligence and machine learning technology to power several features. We believe in transparency about how AI is used in our service.

1.1 AI Provider

We use Anthropic's Claude API for AI processing. Anthropic is a leading AI safety company. As of our last review (January 2026), Anthropic's API terms state that data sent via API is not used to train their models.

1.2 Features Using AI

Feature	What AI Does
Receipt Parsing	Extracts vendor, date, invoice number, line items, prices from receipt images and PDFs
Project Matching	Suggests which project a receipt belongs to based on PO numbers, addresses, colors
Photo OCR	Reads text from photos of receipts taken with phone cameras
Estimate Parsing	Extracts materials, quantities, and costs from QuickBooks estimate PDFs
Surface Detection	Identifies surfaces (walls, trim, ceilings) from work descriptions
Auto-Assignment	Automatically assigns receipts to projects when confidence exceeds 95% threshold

2. What AI Does Well

Our AI features are designed to:

Save time by automating data entry from receipts
Reduce errors compared to manual transcription
Suggest matches based on multiple signals
Extract structure from unstructured documents

In testing, our receipt parser achieves approximately:

95% accuracy for Sherwin-Williams receipts
93% accuracy for JC Licht receipts

These are general benchmarks. Your results may vary based on receipt quality, format variations, and other factors.

3. What AI Cannot Do

AI cannot:

Guarantee accuracy — Every AI output may contain errors
Understand context you haven't provided — AI only knows what's in the document
Replace your expertise — You know your business; AI doesn't
Make business decisions — Bidding, pricing, and customer commitments are your responsibility
Verify physical conditions — AI can't see what's actually on the job site
Detect fraud — AI processes what it's given; it can't verify authenticity

4. Known Limitations

4.1 Receipt Parsing Limitations

Handwritten receipts may not parse correctly
Unusual receipt formats may cause extraction errors
Faded or damaged receipts may be partially unreadable
Multi-page invoices require clear page separation
Non-English text may not be supported

4.2 Matching Limitations

Projects with similar addresses may be confused
Missing or incorrect PO numbers affect matching accuracy
New customers without history have less context for matching
Color name variations (e.g., "SW 7006" vs "Extra White") may not always match

4.3 OCR Limitations

Blurry or poorly-lit photos produce worse results
Glare, shadows, and angles affect accuracy
Small text may be misread
Thermal receipt paper that has faded may be unreadable

5. Your Verification Responsibility

⚠️ IMPORTANT: You are responsible for verifying all AI-generated content before relying on it for business purposes.

5.1 Before Using AI Outputs

Always verify:

[ ] Parsed amounts match the source document
[ ] Line items are complete and accurate
[ ] Project assignments are correct
[ ] Color codes and names are accurate
[ ] Quantities and units are correct

5.2 Manual Correction Features

PaintLogic provides manual correction features specifically because AI isn't perfect:

Edit receipt headers (vendor, date, invoice number)
Edit line items (products, quantities, prices)
Manually assign receipts to projects
Override AI suggestions

Use these features. When you spot an error, correct it.

5.3 High-Stakes Decisions

For high-stakes decisions (customer bids, contract amounts, payment disputes), we strongly recommend:

Verifying AI outputs against source documents
Using manual confirmation for critical data
Maintaining independent records

6. Confidence Scores

Some features display confidence scores (e.g., "95% confidence"). Important context:

6.1 What Confidence Means

Confidence scores are heuristic estimates of how certain the AI is about its output. They are NOT:

Probabilities of correctness
Guarantees of accuracy
Statistical confidence intervals

6.2 High Confidence ≠ Correct

A "95% confidence" result can still be wrong. The confidence score reflects the AI's internal assessment, not objective truth.

6.3 Auto-Assignment Threshold

PaintLogic auto-assigns receipts to projects when confidence exceeds 95%. This threshold was chosen to balance convenience with accuracy, but auto-assigned receipts should still be reviewed.

7. Limitation of Liability for AI Features

7.1 No Warranty

AI features are provided "AS IS" without any warranty of accuracy, completeness, or fitness for any purpose.

7.2 Exclusion of Damages

PaintLogic is not liable for any damages arising from AI-generated content, including but not limited to:

Financial losses from incorrect parsed amounts
Underbids or overbids based on AI estimates
Customer disputes arising from AI errors
Project losses from misassigned receipts
Tax or accounting issues from incorrect data
Any indirect, consequential, or punitive damages

7.3 Your Acknowledgment

By using AI-powered features, you acknowledge that:

AI outputs are estimates and suggestions
You are responsible for verification
Business decisions based on AI outputs are your responsibility
PaintLogic's liability is limited as stated in our Terms of Service

8. Data Sent to AI Services

8.1 What We Send

When you use AI features, we send relevant data to Anthropic's Claude API:

Receipt images and extracted text
Project information for matching context
Work descriptions for surface detection

8.2 What We Don't Send

We do not send to AI services:

Your password or authentication credentials
Payment card numbers
Social Security numbers
Bank account numbers

8.3 Anthropic's Data Handling

Per Anthropic's API terms (as of our last review):

API data is not used to train AI models
Data is processed to provide the service
Anthropic maintains SOC 2 Type II certification

For current terms, see Anthropic's website.

9. Feedback and Error Reporting

9.1 Help Us Improve

If you encounter AI errors, please report them:

Use the thumbs-down feedback button in the interface
Contact support@paint-logic.com with details

Error reports help us improve accuracy for everyone.

9.2 What to Include

When reporting errors, include:

What the AI produced vs. what it should have been
The source document (if possible)
Any patterns you've noticed

10. Updates to AI Features

10.1 Continuous Improvement

We regularly update our AI features to improve accuracy. These updates may change:

Parsing accuracy (hopefully improving)
Supported receipt formats
Confidence scoring methods
Auto-assignment behavior

10.2 Notification

Material changes to AI features will be communicated through our changelog or in-app notifications.

11. Questions

Technical questions about AI features:
support@paint-logic.com

Legal questions about this disclaimer:
legal@paint-logic.com

Privacy questions about AI data processing:
privacy@paint-logic.com

Last reviewed: [DATE]

Attorney Review Status: 🟡 YELLOW — AI liability is an evolving area; recommend attorney review