WHAT
The Chat pattern allows users to express their intent conversationally through an open input interface. Instead of configuring filters, forms, or tools manually, users describe what they want using natural language, voice, or artefacts such as images or links.
The system analyzes the input using an AI model, generates a response, and allows the user to refine the result iteratively through follow-up messages.
In geospatial applications, the output often includes both textual explanations and spatial results such as maps, highlighted locations, or data visualizations.
WHY
Chat provides a flexible way for users to communicate complex intent without navigating structured interfaces. It is especially useful when the task is exploratory or ambiguous, the user does not know which tools to use, or the system supports many possible actions.
Another major advantage of chat-based input is that interactions benefit from iterative refinement in a conversational setting. Chat shifts the interaction from tool-driven to intent-driven.
WHEN
While it’s widely accepted that manual prompt formulation is difficult, time consuming and error prone, there are situations when an open input is beneficial. Use AI Chat in the following situations:
- The task is exploratory such as data exploration.
- The user intent is complex or unclear and chat can provide decision support.
- The system supports many possible actions, for instance during discovery.
- Flexibility is more important than speed.
- Iterative refinement is expected.
Keep in mind that people have a hard time expressing their intent well. Chat moves the burden of articulating this intent efficiently on the user using. Therefore, avoid using chat in the following situations:
- The task is simple and repeatable.
- Speed and efficiency are critical.
- The intent is already well-defined.
- Structured input is faster and clearer.
HOW
AI Chat enables users to interact with geospatial systems conversationally, allowing flexible expression of intent and iterative refinement of results. It is powerful for exploration and discovery, but should complement – not replace – structured interaction patterns. Chat follows the following conversational loop: Input – Analysis – Output – Refinement.
Input
Chat requires the user to provide context through open input such as text, voice, links, or uploading artefacts such as images or documents. This form of input is often referred to as “prompting”. The quality of the output often varies on the prompting instructions and provided details. A prompt example of “Show areas in California with high wildfire risk near dense population” may result in different map styles while adding “as a heatmap” will return a heatmap style map. The example prompt also includes the location context, “in California”.
Analysis
After the input was submitted, the AI model interprets the request by extracting user intent, any relevant context, and information about the location. If any of this information is missing, the model may report back with a follow-up questions. The system may also combine this information with other context such as current map extent, device location, user profile, and available datasets.
Output
Once the system has calculated an answer, it returns the results in the form of a text explanation, typically paragraphs or bullet points. In map-based apps, the map may become part of the response suggesting datasets to be manually added to the map by the user. Automatically adding responses to the map is acceptable but requires a level of certainty, paired with the ability to be changed, removed, or overridden.
Refinement
The user enters an iterative loop to refine the result through follow-up messages. This dialog continues until the desired outcome is reached. An example of a follow-up message is: “Only show areas within 10 miles of hospitals.
Avoid the following situations:
- Providing an empty chat without guidance or examples.
- Forcing users to use chat when structured tools would be faster.
- Not showing which location the system is using.
EXAMPLE
A geospatial datastore requires an interface to find datasets. AI Chat helps users ask conversational questions to explore available data and visualize it on a map.