WHAT
AI Prompt allows users to express their intent through a single input that triggers an AI-powered response. Unlike AI Chat, which is conversational and iterative, AI Prompt is query-based and typically involves a one-time request followed by a result.
The user provides context either by entering a prompt manually through open input or by clicking a button that contains pre-defined context. This approach is often superior to AI Chat because it moves the burden of articulating intent efficiently from the user to the system.
After the prompt is submitted, the AI model analyzes the request and returns a response. In geospatial applications, the output is typically displayed both spatially on the map and textually in a card, panel, or overlay. The textual output may include explanations, summaries, lists, or tables.
Prompt interactions are discrete and transactional, similar to running a query.
WHY
The AI Prompt pattern provides a fast and efficient way to execute AI-powered queries without requiring a conversational interface. It reduces the interaction overhead compared to chat by eliminating the need for conversational back-and-forth. It also enables systems to offer guided AI functionality through pre-defined prompts, making AI accessible without requiring users to formulate their own instructions.
WHEN
The main characteristic of AI Prompt is that the interaction resembles running a query or analysis rather than a conversation. It is especially effective when the question is well-defined and can be expressed in a single query such as “dangerous intersections”. Speed and efficiency are important. Users typically expect an immediate result such as highlighting the dangerous intersections in the map.
Avoid using AI Prompt when the primary task requires iterative refinement, the users’ intent is unclear or evolving, or the workflow benefits from conversational clarification. In these cases, AI Chat provides a better experience.
HOW
AI Prompt enables users to submit a single request that is interpreted and executed by an AI model. The system then returns results visually in a map and additionally in an alternative format such as text or a list.
The workflow follows this interaction flow: Input → Analysis → Output
Input:
The user provides context through a single prompt. While open text can be used to type a questions, it’s better to provide pre-defined suggestions in the form of links or buttons that operate on the current map extent similar to Search this area pattern. The button may be labeled: “Show dangerous intersections” and uses the current map extent as spatial context.
Suggestion may also be pre-cursors to further prompts, often resembling a parent-child relationship or a hierarchy of categories. An example could be a list of buttons that show business types such as “Food”, “Health Care”, “Retail”. Clicking one of these buttons will reveal a new set of buttons to select from, now listing all business types within that selected category.
Pre-defined prompts reduce cognitive effort and improve reliability by guiding users toward supported queries.
Users can specify the spatial context in the following ways:
- Explicit in prompt
- Pre-defined by system
- Inferred from current map extent
- Inferred from app context (selected feature, current project area, user profile)
The system should always make the location context visible to the user.
Analysis:
After submission, the AI model interpret the prompt by extracting the intent, context, and location and returns the output.
Output:
The system returns results in spatial form in the map. Spatial output may include adding layers to the map, highlighting features, or displaying visualizations such as Heat map, Choropleth map, or Hexbin map.
Optionally, textual results may be returned in a side panel, card, pop-up, or map overlay. Textual output may include summaries, lists, Location list, tables, or other explanations. The map and textual output together form a complete response.
Users should always be able to modify, remove, or override AI-generated results.
EXAMPLE
A geospatial app includes a button labeled: “Show recent earthquakes in the current map extent.” When the user clicks the button, the system sends a pre-defined prompt to the AI model. The model identifies relevant datasets, adds earthquake locations to the map, and displays a summary table in a side panel.
The interaction behaves like an intelligent query rather than a conversation.