Ziezodan

The challenge

Ziezodan has a strong structure for repair reports via a decision tree. The catch: for tenants it feels like a lot of clicking, and the real context (description and photo) only comes right at the end. As a result, reports tend to be generic and less useful for follow-up.

The solution

We designed an AI-driven intake that supplements or partly replaces the decision tree. The tenant starts straight away with a short description and can optionally add a photo. Then:

• AI extracts the room, element and problem from the text (NLP)

• The photo helps to confirm or complete this (image recognition)

• If something is missing, the AI asks targeted follow-up questions (“Which room is this in?”)

• The report is mapped to the exact JSON values from the existing structure, and only then forwarded

What it delivers

• Faster, more natural reporting for tenants

• More consistent reports with more context

• Data that plugs directly into the existing JSON hierarchy, with no changes to the downstream process