Photo-to-Quote: How AI Vision Turns Furniture Photos into Moving Estimates
- AI Vision
- Chatbots
- Case Study
Moving companies have a slow, expensive problem: to quote a job accurately, someone has to see what is being moved. Phone calls, video surveys, in-person visits. For Moovez and Quotely I replaced that with a flow where the customer simply uploads photos and gets an estimate in seconds.
The pipeline
- Upload: the customer adds photos of each room or item from the booking wizard.
- Detect: a vision model lists every movable item it can see (sofa, 3-seater; fridge, double door; boxes, approx. 6).
- Estimate: each item is mapped to typical dimensions, volume and weight, with confidence scores.
- Price: volume, distance, vehicle type and labour go into the pricing engine.
- Review: the customer sees the item list, can add or remove items, and confirms.
That last step matters. The AI is a fast first draft, not the final word. Letting the customer correct the list keeps quotes accurate and keeps trust high.
Prompting a vision model for structured output
Free-text answers are useless to a pricing engine. I ask the model for strict JSON and validate it before anything else runs:
{
"items": [
{ "name": "Sofa", "variant": "3-seater", "quantity": 1, "confidence": 0.93 },
{ "name": "Refrigerator", "variant": "double door", "quantity": 1, "confidence": 0.88 }
]
}
Anything below a confidence threshold is flagged for the customer to confirm instead of being silently added.
Lessons from production
- Bad photos are the norm. Dark rooms, half-visible items, clutter. Guide the customer with examples and let them retake.
- Cache aggressively. The same item photo should never be analysed twice.
- Log every correction. When customers edit the AI's list, that data shows you exactly where the model struggles.
- Keep the human option. Some moves are unusual. A "request a manual quote" button saves everyone frustration.
The result
Quotes that took hours of back-and-forth now take seconds, and movers arrive with an accurate inventory instead of surprises. The same pipeline now powers multiple companies on Quotely's multi-tenant platform.
Want to add photo-based estimation or another AI vision feature to your product? Get in touch.