information
This guideline focuses on guided prompts within dialogs. Variants for popovers and side panels are not included. Product teams may adapt the example provided to fit their use cases.

Intro

Guided prompts provide structured input fields and controls that help users instruct the generative AI model on the desired output without having to write descriptive [external_only]custom prompts[/external_only][internal_only]custom prompts[/internal_only]. Guided prompts are useful when users want to specify attributes, styles, or criteria for the generated content, such as length or language.

Users can provide specific details through form fields and guided UI elements. The system combines these inputs into a backend prompt, offering control and precision in generating output that aligns with user preferences.

[internal_only]For more information on all types of prompt patterns, see Designing Effective AI Prompts.[/internal_only]

Image shows the guided prompts pattern in a dialog with form inputs and options for setting AI generation parameters. The form inputs include a select dropdown option for text structure, with the 'paragraph' option selected; a range slider selector for adjusting length, with the range between 250 and 750 characters selected; a select dropdown option for language, with 'English' selected; and a segmented button for changing the tone of the content, with the middle option 'neutral' selected. The dialog options can be confirmed using the primary AI button labeled 'Apply' or the tertiary 'Cancel' button.

Guided prompt in a dialog

When to Use

2
do
false

Use guided prompts when:

  • Users typically submit a limited set of queries.
  • The system supports only a specific set of actions.
  • Users lack experience in writing prompts.
  • Users need consistent and predictable outcomes.
  • Users need ways to direct the AI model’s output.
dont
false

Don’t use guided prompts when:

  • Users need a high degree of flexibility and control over the outcome. Use custom prompts instead.
  • Tasks are repetitive or common within a workflow. Use quick prompts instead.
  • The use case requires only one or a few specific actions. Use quick prompts instead.
  • User intents or desired outcomes are uncertain or undefined. Use custom prompts instead.
  • Tasks are complex or multi-step, requiring nuanced input. Use custom prompts instead.

Anatomy

The guided prompt pattern uses a variety of components to enable new AI-specific interactions. In this example, it is based on a dialog container that can be triggered via an AI button or AI menu button. Any container or input component available in the design system can be selected and configured by product teams to afford the options provided in the guided prompt. Examples are: label, select, range slider, and segmented button. Refer to each component guideline when adopting it in your use case.

  1. Container: A guided prompt is always based on a container, which provides the structure for content input.
  2. Content: The content consists of a selection of form input components. The structure and visual design depend on the respective guidelines.

    Choose input components based on the use case. Select components appropriate to the context and output, and avoid using dropdown menus as the default for all inputs.
  3. Action: The action is implemented using the AI button or AI menu button.

    The action applies the chosen prompt criteria and ensures AI processing only begins with user confirmation.

    This prevents automatic regeneration or loss of content, avoiding unintended AI modifications.

    The container includes an explicit cancel option as a safeguard, allowing users to explore guided prompt options and exit at any point, with clear control over the process.

    Action placement follows the general guideline for the container. Wording should be ultimately defined depending on the use case[internal_only] and in alignment with user assistance[/internal_only]. For more guidance on terminology, refer to the quick prompts or AI writing assistant guideline.

Image shows a schematic representation of the guided prompt pattern in a dialog, highlighting the container, the form and inputs in the content section, and the action options to confirm and apply the prompt at the footer.

Anatomy of a guided prompt in a dialog container

Behavior and Interaction

The guided prompt flow follows a standard interaction pattern based on the chosen container. In this example, users open the dialog, adjust criteria, and confirm to generate content.

Opening a dialog

Clicking the AI button opens a menu with actions (for example, Compose Text) that can trigger the guided prompt.

Image shows a schematic of an object page floor plan highlighting the guided prompt secondary AI button labelled “Compose Text” applied on the top right of a text area contained in a form.

Guided prompt button applied to a form input in an object page Floorplan schematic

Adjusting criteria

The guided prompt dialog opens with controls to adjust criteria and refine the generated content.

Image shows the guided prompt dialog overlayed in an object page floor plan schematic. The guided prompt parameter controls are displayed under the title 'Text Composer,' with confirmation buttons in the footer: a primary AI button labeled 'Apply' and a tertiary button labeled 'Cancel.' The example includes parameter options for Structure, Adjust Length (in characters), Language, and Change Tone.

Guided prompt in a dialog

Generating content

Clicking the confirmation action starts content generation based on the specified criteria. As the generation process begins, a busy state is displayed, and the AI button changes into the Stop Generating button, allowing users to stop generation at any time.

Image shows a schematic of an object page floor plan showing a text area with a busy indicator, after the selected prompt parameters were applied and the content generation process has begun. The guided prompt AI button has transitioned to a 'Stop Generation' button, enabling users to halt the process.

Busy state behavior during the generative AI process

Refer to the AI writing assistant and regenerate guidelines for how to handle warnings about overwriting content when versioning is not supported.

Refer to the busy indicator guideline for when and how to display the indicator during content generation.

After generation

The AI-generated content appears in the respective field, and the AI button transitions back to the default state.

Image shows a schematic of an object page floor plan showing a focused text area with AI-generated content. The 'Stop Generation' button has transitioned back to guided prompt AI button labeled ‘Compose Text’.

Content generated with guided prompt

Handling errors

If the guided prompt is interrupted or fails, follow the guidelines for error messages in the message handling guidelines.

For a consistent user experience, we suggest using the following error message:

Something went wrong while generating your content. Please try again.

Image shows a dialog containing a text area in the error value state and a value state message below. At the top right of the text area there's a ”Compose Text” secondary AI button.

Error handling for content generation using guided prompts

Combining with quick prompts

The AI menu button can be used to combine the easy access for predefined actions of quick prompts with the structured controls of guided prompts.

Add a meaningful action to trigger the guided prompt dialog. In the example, the menu item is labeled Compose Text, but the final wording should be [internal_only]agreed with your user assistance developer and [/internal_only]tailored to the use case.

For more information, see the guidelines for quick prompts guidelines and grouping AI actions.

Animated GIF shows a schematic page containing a text area with some content and a quick prompt menu button labeled 'Revise.' On click, the AI menu appears and reveals quick prompts options like 'Regenerate' and 'Translate.' The last option, identified under a line separator, triggers the guided prompts dialog and is labeled 'Compose Text.' A dialog titled 'Text Composer' overlays the center of the screen and presents the guided prompt parameters controls. The 'Text Length' selector is adjusted to between 250 and 500 characters, the parameters are confirmed using the primary AI button labeled 'Apply,' and the dialog closes. The text area now shows a busy indicator and a 'Stop generating' button on top right, allowing the AI generation process to be stopped. After a delay, the generation process continues, transitioning to text streaming, which replaces the busy indicator and all the original content. Finally, the 'Stop Generating' button transitions back to the AI menu button labeled 'Revise'.

AI menu for quick prompts, extended with a guided prompt option labeled ‘Compose Text’

Recommendations

Example scenario: Text composition

One of the most common scenarios for using guided prompts is text composition. Since the guided prompt pattern reuses existing components, you must follow the guidelines for the chosen components.

The following are some examples of prompt criteria that could be used in the context of text composition.

Language

Image shows a dropdown select control labeled 'Language' with the option 'English' selected.

Select component used to provide language options

Tone of voice

Image shows a segmented button control labeled 'Tone of Voice' with the options 'Casual', 'Neutral' and 'Formal'. The 'Neutral' option is selected.

Segmented buttons used to show tone of voice options

Text length

Image shows a range slider control labeled 'Text Length (in characters)' with the options between 250 and 750 selected.

Range slider used to set the length of generated output

Structure

Image shows a select control labeled 'Structure' with the dropdown showing a list in alphabetic order with the item 'Paragraph' selected. Other items are 'Bulleted List' and 'Long Form Text'.

Select component used to configure the response structure

Application-specific criteria

For scenarios such as generating a customer response or handling different types of support responses, criteria such as Type of Response can be added, and the selected type can be used as a reference for the content generation process.

Image shows a select control labeled 'Type of Response' with the dropdown showing a list in alphabetic order with the item 'Suggest a Call' selected. Other items are 'Empathize with Feedback', 'Propose Solution as a Technical Expert', 'Provide Product/Service Details', and 'Request More Information'.

Select component used to choose a type of response

Terminology

The following guidance is based on the default labels provided in the quick prompts guidelines. We strongly recommend maintaining these labels to ensure consistency and support user familiarity across AI tooling. Adjust the labels only as needed for your specific use case[internal_only] and consult User Assistance for guidance when making changes[/internal_only].

Standard AI action labels for AI text generation and transformation

<div> <div>AI Button Labels</div> <div>Description</div> </div> <div> <div>Compose Text</div> <div>Compose text using predefined criteria to guide the AI generation.</div> </div> <div> <div>Revise</div> <div>Modify existing content or data using AI.</div> </div>
<div> <div>AI Action Labels for Input Components</div> <div>Description</div> </div> <div> <div>Structure</div> <div>Organize information into different text structures like paragraph or list.</div> </div> <div> <div>Text Length</div> <div>Adjust the length of the text.</div> </div> <div> <div>Language</div> <div>Generate content with the selected language.</div> </div> <div> <div>Tone of Voice</div> <div>Adjust the style or emotional quality of text.</div> </div>
guideline

For actions not covered above, apply the following guidelines:

  • Use a verb in the imperative.
  • Keep AI action labels as short as possible while prioritizing clarity for users.
  • Use the same AI action labels consistently.

For more information, see UI Text Guidelines for SAP Fiori Apps.

Guidelines

Implementation

Visual Design

  • Button (visual design)
  • Button (interaction design, web component)

Guidelines

Implementation