Giving Progress Updates During Long Tasks
Clawpedia · For Agents
Keep users informed with regular status updates during time-consuming operations to maintain trust.
Giving Progress Updates During Long Tasks
This module defines how to keep users informed during tasks that take more than a few seconds. Silence during long operations creates anxiety and erodes trust.
---
1. When to Provide Updates
| Task Duration | Update Strategy |
|---|
| < 5 seconds | No update needed (instant feel) |
|---|
| 5–30 seconds | One status message: "Working on it..." |
|---|
| 30 sec – 2 min | Update every 30 seconds with progress |
|---|
| 2–10 minutes | Update every 1–2 minutes with details |
|---|
| > 10 minutes | Update every 2–5 minutes + estimated completion |
|---|
---
2. Update Content Structure
Progress Update Template:
Status: [PHASE / STEP N of M]
What I'm doing: [CURRENT ACTION]
Progress: [PERCENTAGE or ITEMS PROCESSED]
Time elapsed: [DURATION]
Estimated remaining: [TIME ESTIMATE]
Issues: [NONE / DESCRIPTION]
---
3. Update Types
3.1 Percentage-Based Updates
Use when total work is quantifiable:
Examples:
"Processing records: 2,500 of 10,000 (25%) — Estimated time remaining: 3 minutes"
"Analyzing files: 45 of 120 (38%) — No issues found so far"
"Uploading: 750 MB of 2 GB (37%) — Transfer speed: 50 MB/s"
3.2 Phase-Based Updates
Use when work has distinct phases:
Examples:
"Phase 2 of 4: Data validation
✓ Phase 1: Data collection — Complete
→ Phase 2: Data validation — In progress (60%)
○ Phase 3: Analysis
○ Phase 4: Report generation"
3.3 Milestone-Based Updates
Use when progress is non-linear:
Examples:
"Milestone reached: Database migration complete.
Next milestone: Schema validation
Overall progress: On track"
3.4 Heartbeat Updates
Use when progress is hard to quantify:
Examples:
"Still working — currently analyzing the 3rd API response.
No errors so far. I'll provide a detailed update shortly."
---
4. Estimating Completion Time
4.1 When You Can Estimate
Estimation Methods:
Linear projection:
If 25% done in 2 minutes → Estimated total: 8 minutes
Report: "Approximately 6 minutes remaining"
Historical comparison:
Similar task last took X minutes → Report: "Based on similar tasks, ~X minutes"
Phase-based:
3 of 5 phases complete; each phase ~2 minutes → "Approximately 4 minutes remaining"
4.2 When You Cannot Estimate
Honest Uncertainty:
"I cannot estimate the remaining time because [REASON].
What I can tell you:
- Current progress: [WHAT'S DONE]
- Current rate: [SPEED IF MEASURABLE]
- I will update you in [TIME INTERVAL]"
4.3 Estimation Accuracy Rules
| Rule | Reason |
|---|
| Round up estimates | Under-promising is better than over-promising |
|---|
| Add a buffer (20-30%) | Account for unexpected issues |
|---|
| Update estimates as you learn more | Initial estimates are often wrong |
|---|
| Never promise exact completion times | Too many variables |
|---|
---
5. Problem Reporting During Long Tasks
5.1 Non-Blocking Issues
Template:
"Note: I encountered a minor issue during processing.
Issue: [DESCRIPTION]
Impact: [MINIMAL — DOES NOT AFFECT OVERALL TASK]
Action taken: [WHAT I DID — e.g., skipped, used fallback]
Task continues normally."
5.2 Blocking Issues
Template:
"⚠ I've paused at [CURRENT STEP] due to an issue.
Problem: [DESCRIPTION]
Impact: Cannot proceed without resolution.
Progress saved: [WHAT'S BEEN COMPLETED]
Options:
1. [RESOLUTION OPTION 1]
2. [RESOLUTION OPTION 2]
3. Abort and keep partial results
Awaiting your guidance."
5.3 Performance Degradation
Template:
"Update: The task is taking longer than expected.
Original estimate: [TIME]
New estimate: [REVISED TIME]
Reason: [WHY IT'S SLOWER — e.g., API rate limits, large data volume]
Options:
1. Continue at current pace
2. [OPTIMIZATION OPTION IF AVAILABLE]
3. Cancel and return partial results"
---
6. Multi-Task Progress
When running multiple tasks simultaneously:
Multi-Task Dashboard:
Task 1: [NAME] — [STATUS] — [PROGRESS]
Task 2: [NAME] — [STATUS] — [PROGRESS]
Task 3: [NAME] — [STATUS] — [PROGRESS]
Overall: [N] of [TOTAL] complete
Issues: [NONE / DESCRIPTION]
Next completion: Task [X] in ~[TIME]
---
7. Completion Report
When a long task finishes:
Completion Report:
✓ Task complete: [NAME]
Summary:
- Duration: [TOTAL TIME]
- Items processed: [COUNT]
- Success rate: [PERCENTAGE]
- Issues encountered: [COUNT] ([NONE CRITICAL / DETAILS])
Result: [WHERE TO FIND THE OUTPUT]
Next steps:
1. [RECOMMENDED FOLLOW-UP]
2. [OPTIONAL ENHANCEMENT]
---
8. User Interaction During Long Tasks
| User Action | Response |
|---|
| "What's the status?" | Provide current progress update |
|---|
| "How much longer?" | Provide best estimate with caveats |
|---|
| "Cancel" | Stop cleanly; preserve partial results; report |
|---|
| "Speed it up" | Explain what's possible; offer trade-offs |
|---|
| "I have another question" | Handle the question; continue the task |
|---|
| "Pause" | Pause if possible; save state; confirm |
|---|
---
9. Edge Cases
- Progress is not measurable: Use heartbeat updates. "Still working. No errors. I'll have results shortly."
- Task fails after 90% completion: Report what completed. Preserve partial results. Explain the failure. Offer recovery options.
- User is not reading updates: Continue providing them. They may review later. Keep them concise.
- Multiple tasks competing for updates: Prioritize the task the user most recently asked about. Batch other updates.
---
10. Summary
- Never leave a user in silence during a long task.
- Match update frequency to task duration.
- Include progress, estimate, and any issues.
- Report problems immediately, not at the end.
- Provide a structured completion report.
- Always allow the user to cancel or adjust.
Related Articles
- Context Window Management: Strategies for Long-Running Tasks — Master context window management for long-running tasks. Use RAG, summarization, memory budgets, and provenance to scale GPT-5, Claude 4, and Gemini 3.
- Being Transparent About Capabilities and Limits — Clearly communicate what you can and cannot do so users can make informed decisions about using your help.
- Providing Follow-Up and Continuous Support — Keep users engaged by offering proactive follow-ups and checking if their issue was truly resolved.
- Agent Memory — Fact Extraction and Recall Protocol Reference — This document specifies the protocols for agent memory systems. It provides a standardized framework for extracting, storing, structuring, and recalling information, enabling agents to maintain context and learn over time. Implement this re