AI Capabilities
AI features built into Luma — from voice input to financial intelligence.
Voice Agent
Available on: Mobile (long-press FAB button)
A conversational AI assistant that lets field workers interact with Luma by speaking instead of typing. Designed for hands that are dirty, wet, or holding tools.
How it works
- Long-press the floating action button
- Full-screen voice interface opens
- Speak naturally in Swedish
- Luma transcribes, processes, and responds
- Responses are read aloud via text-to-speech
Voice commands
| Command | Example | What Luma does |
|---|---|---|
| Create customer | "Lägg till kund Andersson Bygg" | Creates a new customer record |
| Create job | "Skapa ett jobb åt Eriksson, byta kran i köket" | Creates a work order |
| Log time | "Jag jobbade 3 timmar hos Nilsson idag" | Creates a time entry |
| Set reminder | "Påminn mig att ringa leverantören imorgon" | Creates a reminder |
Technical details
- Language: Swedish (full conversation support)
- Transcription: Real-time speech-to-text
- AI model: Claude (Anthropic) for reasoning and tool use
- TTS: Text-to-speech for responses
- Context: Multi-turn conversations — remembers what was said earlier
- Interface: Full-screen bottom sheet with chat bubble UI
Invoice AI
Available on: Mobile, Web
AI-powered analysis of invoice drafts before sending to the customer.
Features
| Feature | Description |
|---|---|
| Margin analysis | Calculates profit margin per line item and total |
| Pricing suggestions | Recommends price adjustments based on market data |
| Profit optimization | Suggests ways to improve profitability |
| Description improvement | Rewrites line item descriptions to be professional and clear |
| Risk analysis | Flags potential issues (unusually low prices, missing items) |
How it works
- Generate invoice from a completed job
- Review the draft in the invoice builder
- Click "AI Analysis" to get suggestions
- Apply suggestions one-by-one or all at once
- Review and send
AI Copilot
Available on: Web (sidebar panel)
A persistent AI chat interface in the web dashboard. Provides business guidance and answers questions about the user's data.
Capabilities
- Answer questions about jobs, customers, and invoices
- Suggest next actions based on current state
- Explain financial concepts in plain language
- Help with decision-making
- Learn from user behavior patterns
Business Suggestions
Available on: Mobile (My Day), Web (Dashboard)
Proactive AI-generated suggestions based on the user's business activity.
Types of suggestions
| Suggestion | Trigger |
|---|---|
| "Invoice this job" | Job completed but not yet invoiced |
| "Follow up on payment" | Invoice sent but overdue |
| "Log your time" | Calendar events without time entries |
| "Complete this reminder" | Overdue reminders |
| "Review these expenses" | Unlinked expenses |
How it works
- Behavioral AI analyzes patterns
- Suggestions ranked by urgency and impact
- Displayed on dashboard and My Day screen
- One-tap to take action
Receipt OCR
Available on: Mobile
AI-powered receipt scanning and data extraction.
What it extracts
| Field | Description |
|---|---|
| Vendor name | The business that issued the receipt |
| Total amount | The total including VAT |
| VAT amount | Separated VAT amount |
| Date | Receipt date |
| Line items | Individual items when readable |
How it works
- Take a photo of the receipt
- OCR extracts text from the image
- AI parses structured data from the text
- User reviews and corrects any errors
- Expense is created and categorized
AI Data Import
Available on: Web
AI-assisted data migration from other systems.
Features
| Feature | Description |
|---|---|
| Schema discovery | Automatically detects columns and data types from uploaded files |
| Field mapping suggestions | AI recommends how source columns map to Luma fields |
| Data validation | Flags potential issues before import |
| Preset templates | Pre-configured mappings for common systems |
AI Architecture
| Component | Technology |
|---|---|
| Reasoning & Tool Use | Anthropic Claude |
| Text Generation | OpenAI + Anthropic (via Vercel AI SDK) |
| Speech-to-Text | Expo Audio + transcription service |
| Text-to-Speech | Expo Speech (native) |
| Embeddings | pgvector (PostgreSQL) |
| Behavioral Learning | Custom pattern detection |
Guardrails
- AI quota management per organization
- Usage tracking (tokens, cost, execution time)
- Audit logging of all AI operations
- Rate limiting to prevent abuse
- GDPR-compliant data handling — no data sent outside EU