Prompt Patterns
A prompt pattern is a reusable template that tells AI Assist exactly what kind of output you need. Instead of writing a new prompt every time, you select a pattern that includes the structure, constraints, and context AI Assist needs to produce consistent results.
Patterns matter because prompt quality directly determines suggestion quality. Teams that standardize their prompts get more predictable outputs and spend less time refining individual requests.
Why prompt design matters for prototyping
The same AI model produces drastically different results depending on how you ask. A prompt like "create a settings page" yields a generic layout. A prompt like "create a settings page for a B2B SaaS admin with sections for billing, team members, and integrations, using our compact table style" yields something immediately useful.
Without prompt patterns, each team member develops their own prompting style, leading to inconsistent outputs across the project. One person gets clean layouts while another gets cluttered ones, not because of skill differences but because of prompt differences.
Prompt patterns capture the team's collective learning about what works. When someone discovers that specifying the user role and primary action in every prompt improves results, that discovery becomes embedded in the pattern for everyone to use.
Building effective prompt patterns
- Start by analyzing your five most successful AI Assist interactions. Identify what made those prompts effective — specificity about audience, clear output format, explicit constraints, or contextual background.
- Extract the reusable structure from those prompts into a template with placeholder variables. A good pattern has fixed sections (context, constraints, output format) and variable sections (specific feature, user role, content type).
- Test the pattern with at least three different variable inputs to verify it produces consistently useful results. Adjust the fixed sections if outputs are inconsistent.
- Add the pattern to your workspace prompt library with a descriptive name, example usage, and notes about when to use it versus alternative patterns.
- Set pattern defaults that align with your design system. If your team always uses a specific grid, spacing scale, or typography hierarchy, encode those constraints in the pattern.
- Review and update patterns quarterly or whenever the team notices declining suggestion quality. Prompt effectiveness can degrade as models update.
Prompt pattern anti-patterns
- Creating patterns that are too generic to constrain the output. A pattern that says "create a [page type]" adds no value over a freeform prompt.
- Building patterns with so many constraints that AI Assist cannot produce useful variations. Leave room for creative interpretation in areas where flexibility helps.
- Storing patterns without example outputs. Team members need to see what a pattern produces before they trust it enough to use it.
- Maintaining a large library of rarely-used patterns. Prune patterns that are used less than once per month to keep the library navigable.
- Writing patterns in abstract language instead of concrete specifications. "Make it user-friendly" is meaningless to the AI; "include progressive disclosure with a visible primary action and collapsed secondary options" is actionable.
Measuring prompt effectiveness
- Output usability rate: The percentage of pattern-generated suggestions that teams use with minor modifications versus complete regeneration. Aim for sixty percent or higher.
- Pattern adoption: How many team members use shared patterns versus writing freeform prompts. Low adoption suggests patterns are hard to find or do not match real needs.
- Time saved per screen: Compare average screen creation time with and without patterns. Effective patterns should reduce creation time by at least thirty percent.
- Consistency score: Rate the visual and structural similarity of outputs generated from the same pattern by different team members. High consistency indicates the pattern is well-specified.
When to use structured prompts
- When multiple team members need to produce similar outputs for the same project, ensuring visual and structural consistency.
- When building prototype variants for A/B testing where the structure should be identical but the content or layout differs in controlled ways.
- During rapid exploration phases where the team needs to generate ten or more screen variations in a single session.
- When handing off prototype creation to a team member who is new to the project and needs guidance on established patterns and conventions.
Key concepts
- Prompt pattern: A reusable template for instructing AI Assist to produce consistent, relevant output across similar tasks. Patterns reduce variation and improve output quality.
- Context window: The amount of surrounding information the AI considers when generating a response. Larger context windows produce more relevant suggestions but may slow response time.
- Output constraint: A rule that limits what the AI can suggest, such as staying within the design system or using only approved terminology.
FAQ
- What makes a good prompt pattern? Specificity, context, and constraints. A good pattern tells the AI exactly what output is needed, provides relevant context about the project, and sets boundaries on what is acceptable.
- How many prompt patterns should a team maintain? Start with three to five patterns for your most common tasks. Add patterns only when you identify repeated prompt-writing that could be standardized.
- Can prompt patterns be shared across teams? Yes. Patterns stored in the workspace library are available to all team members and can be versioned like other shared assets.
Next steps
Pick the prompt pattern that matches your most frequent prototyping task and test it on a current project. Refine the pattern based on output quality, then document the working version as a team reference. Add new patterns as you discover recurring use cases.