← Back to Dashboard

AI Agent Workflow

Understanding how your autonomous marketing operations agent works

1Data Collection Phase
Gather information from all your sources
Extract

Fireflies Meeting Transcripts

Fetches recent meeting transcripts automatically. Saves meeting metadata (title, date, participants, full transcript) to database.

Google Drive Documents

Monitors shared folders for strategy docs, briefs, and planning documents. Extracts text content from Docs and PDFs.

AI Extraction

Uses AI to analyze content and extract: Decisions (key choices made), Action Items (tasks that need doing), Priorities (urgency levels), and Context (background info).

What You'll See:

  • • Meetings saved to database with full transcripts
  • • Decisions logged with titles and descriptions
  • • Action items created with status "pending"
2Approval Phase
Get human approval before creating tasks
Approve

Slack Approval Requests

Sends formatted messages to your Slack channel with all extracted action items. Includes interactive buttons: ✅ Approve, ❌ Reject, 📝 Edit.

Human Decision Point

Team members review the action items, check priorities, verify assignees, and decide what should become Jira tasks.

What You'll See:

  • • Slack message with formatted action item list
  • • Action items marked as "pending_approval" in database
  • • Button clicks update status to "approved" or "rejected"
3Jira Task Creation
Structure approved items into Jira cards
Create

Context/Action/Acceptance Format

Each approved action item becomes a Jira task with: Context (why it matters), Action (what needs to be done), Acceptance Criteria (definition of done).

Parent Tasks & Subtasks

Creates parent epics for major initiatives and breaks them down into actionable subtasks with proper linking.

Priority Mapping

Automatically maps priorities: High → P1, Medium → P2, Low → P3. Assigns to specified team members.

What You'll See:

  • • New Jira issues created with full descriptions
  • • Tasks linked to source meetings/documents
  • • Action items marked as "created_in_jira" with issue keys
  • • Slack notification with links to created tasks
4Synchronization & Reporting
Keep everything in sync and monitor progress
Sync

Jira Status Sync

Continuously checks Jira for status updates (To Do → In Progress → Done). Updates local database to match.

Blocker Detection

AI analyzes task statuses and comments to identify blockers. Sends alerts when tasks are stuck or dependencies aren't met.

Status Reports

Generates executive summaries: key achievements, upcoming milestones, risks/concerns. Sends to Slack with task metrics.

What You'll See:

  • • Database synced with current Jira status
  • • Blocker alerts in Slack when issues detected
  • • Weekly status summaries with progress metrics
Run Modes
The agent supports different execution modes for flexibility

Full Mode

Runs all 4 phases: extraction, approval, task creation, and reporting. Default for manual runs.

Extract Only

Just phase 1 - fetches data and extracts insights without creating tasks. Good for testing.

Sync Only

Just phase 3-4 - syncs Jira statuses and generates reports. Runs automatically on schedule.

Report Only

Just phase 4 - generates status summaries and detects blockers. Useful for quick updates.

Checking Results
Where to find extracted data and created tasks
/data

View all extracted meetings, decisions, and action items in organized tables

Jira Project

See created tasks with full Context/Action/Acceptance format

Slack Channel

Receive approval requests, task notifications, and status updates

Dashboard

Monitor agent runs, view metrics, and see recent activity at a glance