Scheduling And Automation

Choose the right scheduling pattern for recurring Kommon Poll AI workflows.

Scheduling support varies substantially between AI clients. Do not assume that a workflow available in one platform is available in another.

When deterministic delivery, strict retries, or production logging are required, use the Kommon Poll API rather than MCP.

1. Scheduling Patterns

2. Example Schedules

3. Define Each Scheduled Workflow

Before scheduling, define:

• Saved search or searches.
• Reporting period.
• Timezone.
• Filters.
• Output format.
• Recipient or destination.
• Failure behavior.
• Retry behavior.
• Whether human approval is required.
• Whether the AI client can access Kommon Poll tools in scheduled mode.

4. Native Scheduled AI Task Pattern

Use this only where the AI client supports scheduled tasks with Kommon Poll tool access.

Prompt:

Every Monday morning, use Kommon Poll to create a weekly executive brief for the "Brand Monitoring" saved search covering the previous 7 days in Asia/Colombo time. Include material changes, sentiment, reach, interactions, high-influence mentions, risks, opportunities, and recommended actions. State the date range, filters, and limitations.

Before relying on this pattern, verify:

• The client supports scheduling.
• The scheduled mode can use MCP tools.
• The Kommon Poll connection remains authorized.
• The output destination is acceptable.
• Failures are visible to the workflow owner.

5. Agent Or Automation Platform Pattern

Use this when a managed automation layer should trigger the workflow.

Typical flow:

1. Scheduler or event triggers an agent.
2. Agent calls Kommon Poll MCP tools.
3. Agent writes a brief.
4. Human approval is requested if required.
5. The brief is posted, emailed, saved, or routed to another system.
6. Run status is logged.

This pattern is useful when the workflow needs governance or delivery into enterprise systems.

6. API Scheduled Job Pattern

Use the Kommon Poll API when you need:

• Deterministic calls.
• Cron or serverless scheduling.
• Explicit retries and timeouts.
• Full logging.
• Warehouse or BI ingestion.
• Repeatable report generation.
• Strict secret handling.

The API can produce the data payload, and an AI model can still be used downstream for writing or summarization if your architecture allows it.

7. Platform Support Matrix

Fill this table after verifying current platform behavior.