Skip to content

Draft v0.1

Express Intent

Lifecycle Stage 1
This article is part of the AI Interaction Lifecycle framework
Express Intent → Display Results → Refine Output → Take Action → Execute & Monitor → Summary

Overview

Every AI interaction begins with intent.

Before an AI system can provide information, generate content, perform actions, or execute workflows, it must first understand what the user is trying to accomplish. Expressing Intent focuses on the patterns that help users communicate their goals clearly, efficiently, and with confidence.

The quality of the interaction that follows is often determined by the quality of the initial intent. Effective AI experiences help users discover available capabilities, formulate requests, provide context, and refine objectives before the AI begins generating responses or taking action.

Whether the experience is informational, assistive, or fully agentic, helping users express intent effectively is the foundation of successful AI interaction.

Initial Call to Action (CTA)

For many products, the first touchpoint with AI is the Initial Call to Action (CTA): a prominent collection of inputs and actions that allow users to begin interacting with the AI system.

The CTA serves as the primary entry point into the AI experience and establishes expectations for how users can engage with the system.

Depending on the experience, a CTA may include:

  • A prompt input field
  • Suggested prompts
  • Templates
  • Recent activities
  • File upload capabilities
  • Context-aware actions

From this starting point, users work with the AI through prompts, refinements, regenerations, and follow-up interactions to achieve their intended goals.

Supporting Patterns

The following patterns help users formulate requests and discover effective ways of interacting with AI systems.

Predictive Assistance

Predictive Assistance enables AI systems to anticipate user needs and provide proactive guidance before a request is fully formed.

By leveraging historical behavior, contextual signals, and current user activity, the system can suggest likely actions, questions, or next steps.

Predictive assistance reduces friction and helps users move from intent to outcome more efficiently.

Prompt Presets

Prompt Presets provide predefined prompts that help users get started quickly.

These predefined requests:

  • Reduce cognitive effort
  • Eliminate prompt-writing anxiety
  • Demonstrate system capabilities
  • Produce more consistent outcomes

Prompt presets are especially valuable for onboarding new users and introducing unfamiliar features.

Suggestions

Suggestions act as conversation starters that help users understand what the AI can do.

Rather than requiring users to generate requests from scratch, suggestions present examples of useful tasks or questions that can be immediately executed or modified.

Suggestions are frequently displayed:

  • At session start
  • After task completion
  • During periods of inactivity
  • As contextual recommendations

Templates

Templates provide structured starting points for recurring or complex tasks.

They help users:

  • Follow best practices
  • Maintain consistency
  • Reduce learning curves
  • Accelerate task completion

Examples include:

  • Project plans
  • Meeting summaries
  • Status reports
  • Research requests
  • Content briefs

Templates are particularly effective when users need guidance on how to frame a request.

Nudges

Nudges use progressive disclosure to help users discover AI capabilities naturally within a workflow.

Rather than interrupting the user, nudges appear contextually when AI assistance may be helpful.

Examples include:

  • "Would you like AI to summarize this document?"
  • "Generate next steps?"
  • "Draft a response?"
  • "Analyze this data?"

Nudges introduce AI at moments of high relevance and encourage adoption through contextual assistance rather than instruction.

Memory

Memory gives AI systems the ability to retain and reuse information across interactions, creating continuity for the user.

Instead of starting from scratch in every session, the system can recall prior context, preferences, instructions, or facts and apply them to future tasks. This shifts the AI from a transactional tool into a more persistent and personalized assistant.

Global Memory

Information persists across all surfaces and contexts.

Global memory works well for long-term preferences such as:

  • Communication style
  • Preferred tone
  • Language preferences
  • Personal settings
Example:
ChatGPT's cross-conversation memory.

Scoped Memory

Information is retained within a specific workspace, project, or conversation.

Scoped memory helps maintain relevance while preventing unnecessary context from carrying into unrelated tasks.

Example:
Perplexity Spaces.

Ephemeral Memory

Information is retained only for the duration of the current session.

This approach is particularly useful for:

  • Privacy-sensitive scenarios
  • Temporary research
  • Anonymous interactions
  • Exploratory workflows
Example:
Apple Intelligence's session-based context model.

Principles Applied

This stage primarily aligns with the following AI UX principles:

  • Control – Users should be able to define and shape their requests while maintaining ownership of the interaction.

Next Stage: Display Results

Once user intent has been captured, the AI transitions from understanding the request to generating a response.

The next stage in the lifecycle, Display Results, focuses on how AI communicates outputs, explanations, confidence signals, citations, and supporting evidence to help users understand and evaluate generated outcomes.