Patterns

AI design patterns

Our AI patterns are a helpful resource for designing new Iterable AI features or product experiences. While they are not absolute, they play an instrumental role in creating a unified suite of AI-powered solutions.

 

Set clear expectation

  • Clear communication of purpose and value. Detailing relevant inputs as well as likely outcomes for informed decision
  • Any limitations, trade offs, or potential consequences should be straightforward to ensure user confidence and trust

 

An example of how communicating expectations is handled with predictive goals.

Error affordance

  • There will be instances where the AI fails to perform as expected. The errors should be be surfaced for visibility
  • When the AI fails, make sure it is as clear as possible to the user why there is a failure and what the implications of the failure are
  • Even for situations where the cause and reason are unclear, offer the user an option of what to do next

 

Familiarity over novelty

  • Common UX patterns and heuristics should be prioritized over entirely new modes of interaction to ease input and reduce uncertainty or confusion
  • Unique design patterns should be componentized and added to an Iterable AI pattern library for reuse
  • Novelty in presentation and adornment should be considered when there is specific value in differentiating the experience for clarity or branding purposes

 

Automation vs. control

  • The balance between automation and user control should be considered based on the context and scope of the AI experience
  • Automating tasks can help users work more efficiently and effectively, but should always include options for more (or less) control depending on user preference and risk tolerance
  • Lightweight, background tasks may skew toward automation, while those with greater impact will require more explicit user input and configuration

 

Anticipate the unknown

  • There are often “known unknowns” with any AI tool. A robust error framework is critical, but often not enough
  • Designing (and often building) for various edge cases helps hedge the experience against unclear or unexpected issues
  • Giving users control over how to respond to scenarios such as inconclusive data, we can build trust and confidence in the tool