Playground
  • Introduction
  • Components

Next-best action

The system analyses what the actor is doing and recommends the most useful next step.

The most forward-leaning pattern on the assisted task completion spectrum, right at the edge of autonomous action. It works at the workflow level, not at the field or content level, and suggests actions — schedule a call, run tests, send a follow-up — rather than data values or text.

Related patterns

Precedes

  • Cognitive forcing functions — workflow-level recommendations that benefit from forcing functions when stakes are high
  • Action consequences — when accepting a recommendation leads to consequential decisions

Instantiates

  • Suggestion — suggestion is the broader pattern; next-best action specialises it to recommending workflow steps

Enacts

  • Agency — proactive recommendations must preserve actor control
  • Adaptability — recommendations that improve from accepted/rejected signal over time

Complements

  • Assisted task completion — the spectrum this pattern belongs to
  • Bot — powers next-best action via the ambient mode of interaction
  • Embedded intelligence — next-best action as intelligence woven into the workflow
  • Transparent reasoning — explaining the rationale behind recommendations
  • Notification — a common delivery mechanism for next-best action prompts
  • Onboarding — next-best action as guided first-use experience

Alternatives

  • Wizard — the wizard prescribes the sequence; next-best action is the adaptive, context-aware alternative

Related

  • Assistance — Next-best action draws on assistance to decide *what* to suggest at workflow transitions
  • Delegation — Next-best action straddles the suggestion/delegation line — the same recommendation can be presented for the actor to accept or for the agent to perform.