Activity Monitor

The Activity Monitor provides a unified, real-time view of all agent activities across your workspace. From a single screen, you can track task status, review completed work, unblock agents requiring assistance, and maintain complete operational visibility over your AI operations platform.

​

What is the Activity Monitor?

The Activity Monitor is your operations dashboard for managing autonomous AI workflows. It provides:

• Real-time status tracking - See running, completed, and stopped agents at a glance
• Quick task preview - Review task details and outputs without leaving the list view
• Unblocking capability (Enterprise) - Provide guidance to agents that need human input
• Complete execution history - Access past work, decisions, and deliverables
• Operational intelligence (Enterprise) - Understand agent performance and workflow patterns

Think of it as your mission control for AI operations—where you maintain oversight, intervene when needed, and track the autonomous work happening 24/7 across your organization.

​

Key Capabilities

​

1. Unified Task View

All agent activities appear in a single, organized table: Status Indicators:

• ✅ Done - Successfully completed tasks
• 🔄 Running - Agents currently executing
• ⏸️ Stopped - Agents waiting for human input or guidance
• ❌ Failed - Tasks that encountered errors

Visible Information:

• Task name/description
• Current status or completion summary
• Time since created/completed
• Agent owner/creator
• Quick action menu

​

2. Quick Task Preview

Click any task row to see detailed information in the side panel: Overview Section:

• Task title and status
• Completion timestamp
• Full task description

Deliverables Section:

• Files created (Excel, PDFs, reports)
• Documents generated
• Emails sent
• System actions taken
• Links to outputs

Execution Details:

• Step-by-step task progression
• Tools and integrations used
• Data accessed and processed
• Reasoning and decisions made

Quick Actions:

• Copy task
• Share task
• Export outputs
• View full conversation
• Expand to full-page view

​

3. Agent Unblocking

When agents encounter ambiguity or need guidance, they stop and request human input. Unblocking Workflow:

1. Identify stopped agent - Shows “Stopped” status in the list
2. Click task row - Opens detail panel showing where agent is stuck
3. Review agent’s question - Understand what guidance is needed
4. Reply with instruction - Provide clarification or direction
5. Watch status transition - Agent moves to “Running” → processes request → returns to “Done”

Example Unblocking Scenarios:

​

Scenario 1: Ambiguous Request

• Agent stops: “I found 3 different interpretations of ‘executive team.’ Which should I analyze: C-suite only, VPs and above, or entire senior management?”
• You reply: “Focus on C-suite only: CEO, CFO, CTO, COO”
• Agent resumes: Completes analysis with correct scope

​

Scenario 2: Missing Information

• Agent stops: “To complete the compliance review, I need the risk tolerance threshold. Should I use: Conservative (<5% variance), Moderate (<10%), or Aggressive (<15%)?”
• You reply: “Use Conservative threshold (<5%)”
• Agent resumes: Applies threshold and completes review

​

Scenario 3: Decision Point

• Agent stops: “Found 2 critical security vulnerabilities. Should I: 1) Auto-patch immediately, 2) Notify security team first, or 3) Schedule maintenance window?”
• You reply: “Notify security team first, then schedule patch for tonight’s maintenance window”
• Agent resumes: Executes approved workflow

​

4. Filtering & Organization (Enterprise)

Status Filters:

• View only completed tasks
• Show running agents
• Filter stopped agents needing attention
• Display failed tasks requiring intervention

Time Filters:

• Today’s activity
• This week
• This month
• Custom date range

Agent Workspace Filters:

• View tasks from specific agent workspaces
• Filter by agent type (email assistant, research agent, etc.)
• Show only your tasks vs. team tasks

Search:

• Search by task name
• Search by task content
• Search by deliverables
• Search by date range

​

5. Batch Operations

Select multiple tasks for bulk actions:

• Archive completed tasks - Clean up the dashboard
• Export task history - Download execution logs for reporting
• Share multiple results - Send batch of completed work to team
• Compare tasks - Analyze similar workflows for optimization

​

Real-World Use Cases

​

Use Case 1: Daily Operations Review

Morning Routine:

1. Open Agent Activity Monitor
2. Check “Done” section - Review overnight automated work
3. Review deliverables - Verify reports, summaries, data processing completed correctly
4. Check “Stopped” section - Unblock any agents awaiting guidance
5. Monitor “Running” section - See current operations in progress

Result:

• Complete visibility in under 5 minutes
• Proactive issue resolution
• No work blocked by unavailable team members

​

Use Case 2: Customer Support Operations

Scenario: Support team uses agents to triage and draft responses to tickets Monitor View:

• 48 completed tasks - Tickets triaged with draft responses
• 12 running tasks - Current tickets being researched and drafted
• 3 stopped tasks - Require human judgment on edge cases

Workflow: Review stopped agents:

• Agent 1: “Customer requesting refund outside policy window but has extenuating circumstances. Approve exception?”
  • Unblock: “Approve one-time exception, note in CRM”
• Agent 2: “Ticket requires technical escalation. Assign to: DevOps team or Platform team?”
  • Unblock: “Escalate to Platform team, priority: high”
• Agent 3: “Customer sentiment extremely negative. Should I: draft empathetic response, escalate to manager, or offer proactive compensation?”
  • Unblock: “Escalate to manager @sarah, draft empathetic holding response”

Monitor completion - All 3 agents resume and complete within minutes Result:

• 48 tickets handled autonomously
• 3 edge cases resolved with human judgment
• Complete documentation of decisions

​

Use Case 3: Security Operations Center (SOC)

Scenario: Security team uses agents for alert triage and initial investigation Monitor View:

• 120 completed tasks - Alerts triaged, false positives documented
• 8 running tasks - Active investigations in progress
• 2 stopped tasks - Require security engineer assessment

Workflow: Review stopped agents:

• Agent 1: “Detected unusual API access pattern from IP 203.x.x.x. Matches both: legitimate partner traffic AND suspicious reconnaissance behavior. Confidence: 65% benign, 35% malicious. Recommendation?”
  • Engineer unblocks: “Add IP to watchlist, enable enhanced logging, allow traffic but monitor closely”
  • Agent resumes: Implements monitoring, documents decision
• Agent 2: “Found privilege escalation attempt on production database. Automatically blocked access. Should I: 1) Maintain block and alert only, 2) Force password reset for affected users, 3) Initiate full incident response protocol?”
  • Engineer unblocks: “Initiate full incident response, page on-call team”
  • Agent resumes: Executes incident protocol, notifies stakeholders

Result:

• 120 alerts auto-triaged in hours (would take days manually)
• 8 investigations running in parallel
• 2 critical decisions made by security engineer with full context
• Complete audit trail maintained

​

Use Case 4: Financial Close Operations

Scenario: Finance team running month-end close with agent automation Monitor View (1st business day of month, 10 AM): 15 completed tasks:

• Stripe transaction export: ✅ Done
• Salesforce revenue reconciliation: ✅ Done
• QuickBooks invoice matching: ✅ Done
• Variance analysis: ✅ Done (flagged 3 items >10% variance)
• Preliminary close report: ✅ Done

2 stopped tasks:

• Agent 1: “Found $47,000 unreconciled difference in Services revenue. Possible causes: 1) Delayed invoice recording, 2) Revenue recognition timing, 3) Data entry error. Investigation required.”
• Agent 2: “Customer ABC Corp payment received but no open invoice found. Amount: $125,000. Should I: create revenue entry, research contract, or flag for manual review?”

Finance Manager Actions:

1. Click Agent 1 task - Reviews analysis
   • Unblocks: “This is the XYZ Enterprise deal that closed late Q4 but revenue recognized in January. Create accrual entry with ref: Contract #2024-089”
   • Agent resumes: Creates proper accounting entry, updates reconciliation
2. Click Agent 2 task - Reviews payment details
   • Unblocks: “This is prepayment for 2025 subscription renewal. Record as deferred revenue, notify AR team”
   • Agent resumes: Processes correctly, sends notification

Result:

• 90% of close process automated
• Edge cases flagged with context for rapid resolution
• Close completed by noon (previously took 3-5 days)

​

Use Case 5: Research & Analysis Operations

Scenario: Strategy team using agents for competitive intelligence and market research Monitor View:

• Daily Competitor Monitoring - ✅ Done (posted to #competitive-intel)
• Industry News Summary - ✅ Done (25 relevant articles, 5 flagged as high-impact)
• 100 Company Analysis - 🔄 Running (63/100 completed)
• M&A Target Research - ⏸️ Stopped (requires prioritization input)

Analyst Actions: Review completed daily tasks:

1. Click “Daily Competitor Monitoring”
   • See: Competitor X announced new pricing tier
   • Deliverable: Full analysis in Notion + Slack summary
   • Quick action: Share with product team

Monitor running task:

• “100 Company Analysis” - Watch real-time progress
• 63 companies analyzed in parallel
• Estimated completion: 12 minutes

Unblock M&A research:

• Agent stopped: “Found 47 potential acquisition targets matching criteria. Analysis would take ~6 hours for all. Priority approach: 1) Analyze all 47, 2) Filter to top 20 by revenue, 3) Focus on specific geography/vertical?”
• Analyst unblocks: “Filter to top 15 by strategic fit score (high product overlap + strong growth + geographic expansion opportunity)”
• Agent resumes: Narrows focus, completes deep analysis on priority targets

Result:

• Daily intelligence delivered automatically
• Large-scale research completing in background
• Strategic prioritization applied by human
• Team receives actionable insights, not data dumps

​

Benefits

​

1. Complete Operational Visibility

See every agent activity across your workspace. No black box—understand what’s running, what’s done, what needs attention.

​

2. Rapid Issue Resolution

Identify and resolve blocked agents in seconds. Don’t let automated workflows sit idle waiting for simple guidance.

​

3. Quality Assurance

Review completed work to ensure agents are executing correctly. Spot-check outputs, verify reasoning, maintain quality standards.

​

4. Performance Analytics

Track agent execution patterns. Identify workflows that frequently require unblocking. Optimize playbooks based on real operational data.

​

5. Audit Trail

Complete history of every agent action. Who initiated what, when completed, what was delivered. Essential for compliance and governance.

​

6. Team Coordination (Enterprise)

See work across entire team. Understand who’s running what, where coordination is needed, what work is in progress.

​

Best Practices

​

Daily Operations

Morning Review (5 minutes):

• Check “Done” section - Verify overnight automation completed
• Review “Stopped” section - Unblock any waiting agents
• Monitor “Running” section - See current operations

Throughout Day:

• Check monitor when notifications arrive
• Unblock agents promptly (< 30 minutes)
• Review deliverables from completed tasks

End of Day:

• Archive completed tasks from previous days
• Verify no agents left in “Stopped” state overnight
• Review failed tasks, address issues

​

Unblocking Agents

Be Specific:

• ✅ “Use Conservative risk threshold (<5%), analyze C-suite only”
• ❌ “Do what seems reasonable”

Provide Context:

• ✅ “This is prepaid subscription for 2025. Record as deferred revenue per ASC 606”
• ❌ “Handle it appropriately”

Document Decisions:

• Agents learn from your unblocking responses
• Clear guidance improves future autonomous execution
• Builds institutional knowledge into agent behavior

​

Quality Assurance

Spot Check Strategy:

• Review 5-10% of completed tasks randomly
• 100% review of high-risk operations
• Review first few executions of new playbooks

What to Check:

• Accuracy of outputs
• Appropriateness of actions taken
• Quality of reasoning
• Adherence to company policies

​

Common Questions

Q: How long do completed tasks stay in the monitor? A: Indefinitely. Archive manually when no longer needed, or set auto-archive policies (e.g., archive tasks older than 90 days). Q: Can I see tasks from other team members? A: Depends on workspace permissions. Workspace admins see all tasks. Team members see shared agent tasks and their own tasks. Q: What happens if I don’t unblock a stopped agent? A: Agent remains stopped until unblocked. Set notification preferences to alert you immediately when agents stop. Q: Can agents unblock themselves? A: Not by default. However, you can configure fallback behaviors: “If no response within 4 hours, default to conservative approach.” Q: How do I know when an agent needs unblocking? A: Real-time notifications via: browser notification, Slack message, email (configurable per agent/priority). This feature is still in Alpha testing. Q: What’s the difference between “Stopped” and “Failed”? A: Stopped = agent needs human guidance. Failed = agent encountered technical error (API failure, timeout, etc.). Q: Can I export the task history? A: Yes. Export to CSV for analysis, reporting, or auditing. Includes all task metadata, timestamps, and status history. This is for Enterprise only.

---