AI Now Fills Out Entire Report Forms

When Bonn relaunched their request management at anliegen.bonn.de, AI photo recognition was already a headline feature: residents upload a photo and the system suggests the right report category automatically. That was a solid first step. The second step, shipped in the latest versions of Mark-a-Spot, is far more consequential: the AI now fills out entire report forms.

Open311 Service Definitions: The Missing Piece

One of Open311's most powerful features is its service definition system. Each report category can define structured attributes: dropdowns, radio buttons, free-text fields that capture exactly the information municipal staff need to route and resolve an issue. A traffic light report might include "Signal Issue" (button broken, not working, flickering, wrong timing) and "Intersection Type" (pedestrian crossing, vehicle intersection, bicycle signal). A road defect report might ask for defect type and severity.

These service definitions are what separate a vague "something is broken" report from actionable, structured data. But until now, the AI only handled the first half: it suggested the category and a description. The service definition attributes were left for the citizen to fill in manually, one dropdown at a time. Category-specific details that can be freely defined per category were only captured on site by the maintenance crew, leaving them no way to prepare before dispatch.

That friction is now gone.

One Photo, One API Call, Everything Pre-filled

The updated markaspot_vision module changes this. When a citizen uploads a photo, the AI receives not only the image but also the full set of service definitions for all categories, including all their attribute options. It returns category, description, and attribute values in a single response.

Take a real example from Amsterdam: a citizen photographs a damaged traffic signal at a road intersection. The AI recognizes the situation and returns, in a single response, the category "Traffic Signals", sets "Signal Issue" to "Physically Damaged", selects "Vehicle Intersection" as the intersection type, and enters 1 under "Number of Affected Signals". The citizen sees a complete, pre-filled form and only needs to confirm or adjust.

Graceful Degradation Included

Not every category has service definition attributes. For those, the behaviour is unchanged: the AI returns category and description, and the attributes array is simply empty. Nothing breaks, nothing looks different to the citizen.

What This Means in Practice

For citizens, fewer taps and less guesswork. For municipal staff, incoming reports carry more structured, accurate data from the start, which means less manual reclassification and faster routing.

Try It Now

The feature is live on the Amsterdam demo. Upload a photo and watch the form pre-fill automatically. For a full walkthrough with your city's service definitions, request a demo.

Why Structured Data Compounds Over Time

Open311's service definition standard was designed to let municipalities define their own services and categories individually. That pays off immediately: structured reports are easier to route, easier to prioritize, and easier to close.

But the value accumulates. A defect record with type, severity, surface material, and precise coordinates is not just useful today. It feeds dashboards, informs maintenance planning, and becomes training data for the next generation of AI-powered analysis tools. The difference between "something's broken on Main Street" and a structured record is the difference between a ticket that needs human interpretation and one a system can act on directly.

Every report that arrives with properly filled attributes is an investment. Building that habit into the reporting workflow now, through service definitions and AI-assisted form completion, means the data is there when you need it.

Open source, standards-based, built to last.