Pattern Summary

Recommend + Explain is a guidance pattern where the agent suggests a specific action, path, or decision and provides rationale to support it. This helps build user confidence, accelerates decision-making, and makes the agent’s reasoning transparent.

The pattern is especially useful in cases where users need to choose between options or when the agent is proactively guiding toward an outcome.

When to Use It

Use Recommend + Explain when:

  • The agent is making suggestions that affect user outcomes
  • Transparency increases trust or helps with learning
  • Users benefit from understanding trade-offs
  • The recommendation is one of several viable options

Examples include product recommendation agents, workflow automation prompts, system performance tuning, or prioritizing alerts.

How It Works

  1. Recommendation: Agent presents an action or decision suggestion
  2. Rationale: Agent explains why this choice is being made (data, goals, past behavior)
  3. Alternatives (optional): Other options may be acknowledged or shown
  4. User Response: User accepts, modifies, or rejects the recommendation

This pattern works in both proactive and reactive modes.

Fit Assessment

Use this pattern if:

  • The user needs guidance but retains agency
  • Trust, clarity, or education are part of the experience
  • Agent logic or inference is non-obvious

Avoid using it when:

  • The explanation doesn’t add clarity or is overly technical
  • The recommendation is trivial or obvious
  • The user cannot act on the recommendation

Acceptable Dependencies

✅ Clear recommendation logic (rules, ML, heuristics)
✅ UX to support recommendation + rationale display
✅ Logging to track acceptance, edits, and overrides
✅ Optional: confidence scoring or context tags

Unacceptable Dependencies

❌ Unexplainable black-box recommendations
❌ Agent insists without offering a reason
❌ Rationale includes non-actionable or generic justifications

Implementation Starter Guide

  • Make the recommendation a complete sentence, not a label
  • Include clear, human-readable rationale
  • Show when the user’s behavior has influenced the recommendation
  • Offer ways to revise or reject
  • Track how users engage with suggestions

Example: Onboarding Optimization Assistant

The agent notices that users drop off at step 3 of onboarding. It recommends:

"You might consider simplifying Step 3 - 40% of users exit at this point."

[Apply Suggestion] [Edit Flow] [Dismiss]

Strategic Value

  • Makes agent behavior feel intelligent and helpful
  • Encourages learning and trust through transparency
  • Promotes better decisions with less manual research

Recommend + Explain creates a two-way dialogue where agents don’t just act - they persuade and support.

Tags

Pattern Type: Suggestion, Transparency, Decision Support
Scope: User-facing, Multi-modal
Recommended UI Modes: Tooltip, Card, Assistant Panel