Trustworthy & Reliable AI

Action Audit Trail

Provide a timestamped, structured log of every action the agent took - grouped by task, with reversibility status, selective undo, and diff views - so users can review and correct agent behavior after the fact.

What is Action Audit Trail?

After an agent has acted - especially across multiple steps or over extended periods - users need a clear, reviewable record of what happened. This is fundamentally different from traditional undo/redo because agentic actions may span multiple systems, occur asynchronously, and have cascading consequences. A user who discovers their agent has sent 15 emails, rescheduled 3 meetings, and updated a spreadsheet needs to quickly understand what happened, why, and how to reverse specific actions. The Action Audit Trail provides a timestamped log grouped by task with plain-language descriptions, reversibility color-coding (green/amber/red), selective undo capabilities, and before/after diff views for document modifications. This extends source attribution from citing information sources to citing action sources.

Example: Zapier / Make - Automation Execution History

Zapier execution history showing automation steps with success status, timestamps, and expandable detail

Execution history shows every step of an automation with success/failure status, input/output data, and timestamps. Users can inspect individual steps to understand exactly what happened at each stage.

AI Design Prompt

Want to learn more about this pattern?

Explore the full pattern with real-world examples, implementation guidelines, and code samples.

View Full Pattern

Related Prompts from Trustworthy & Reliable AI

Explainable AI (XAI)

Trustworthy & Reliable AI

Design an explainable AI interface that makes decision-making transparent: Create a decision explanation card showing: 1. **Decision Output**: The AI's conclusion or recommendation prominently displayed 2. **Confidence Score**: Visual indicator (progress bar/percentage) showing certainty level 3. **Key Factors**: Top 3-5 factors that influenced the decision with visual weights 4. **Data Sources**: Citations or references to where information came from 5. **Alternative Options**: Other options considered with brief explanations Use visual hierarchy to show the most important factors first. Include an option to "See detailed explanation" for users who want deeper insights.

View Full

Responsible AI Design

Trustworthy & Reliable AI

Design a responsible AI decision interface similar to LinkedIn's AI-powered recommendations or Microsoft's Responsible AI dashboard. Show an AI recommendation card with transparency layers. Include: main decision/recommendation display, expandable 'How this was decided' section showing key factors with visual weights, bias detection indicator (color-coded badge), data source attribution, user control panel with override and feedback buttons, and audit trail timeline. Style: Professional, trustworthy, high-contrast for accessibility. Use blues/greens for trust, clear typography, WCAG AAA compliant. Platform: Web application, responsive design.

View Full

Error Recovery & Graceful Degradation

Trustworthy & Reliable AI

Design an error recovery interface inspired by ChatGPT's 'at capacity' error, GitHub Copilot's offline state, or Grammarly's error handling. Show a friendly error state with clear recovery paths. Include: (1) Prominent but non-alarming error message with warm-colored icon (amber/yellow for capacity/service issues), (2) Plain-language explanation of what happened and why, (3) 'Your work is saved' indicator with green checkmark to reduce user anxiety, (4) 2-3 recovery action buttons clearly labeled (e.g., 'Try Again', 'Wait in Queue', 'Use Offline Mode'), (5) Optional: Queue position counter or estimated wait time, (6) Tip or note about premium/priority access if applicable. Style: Calm, transparent, solution-focused. Use amber/yellow for warnings, green for saved state indicators, black/dark buttons for primary actions. Avoid red unless it's a critical system failure. Platform: Modern web application, responsive design.

View Full

Previous PromptPlan Summary View All Prompts Next PromptEscalation Pathways