Automation tasks

Run MASAIC agents on schedules or outside the chat UI: automation tasks, workflows, patterns, and checklists.

On this page: Tasks list · Workflows · Tasks vs chat · Patterns · Good habits · See also

Automation runs agent workflows without a human typing in the Agent Playground — on a schedule, in batches, or as operational jobs — using the same agents, prompts & resources, and MCP tools you already validated in Chat.

Tasks list

In the sidebar: Workflows → Automation Tasks. You will see configured jobs, status, and actions to create or edit definitions.

Workflows

The same Automation section includes:

Item	Purpose
Workflow definitions	Design multi-step automation flows.
Workflow runs	Inspect execution history and status.

Use tasks for scheduled or batch agent jobs; use workflows when you need orchestrated steps across systems.

When to use tasks vs chat

Situation	Prefer Agent Playground	Prefer automation
Debugging a bad answer	✓
Stakeholder demo in the room	✓
One-off “what if” question	✓
Same job every night / week		✓
Large batch (many rows, same flow)		✓
Must succeed unattended with logging		✓

Common patterns

Digest — Summarize events on a schedule; deliver via your integrations.
Batch enrichment — Classify or extract fields for many records with one flow.
Health checks — Periodic synthetic runs proving agents and tools still respond.

Name tasks after the business outcome (“Weekly support summary”), not only the underlying model.

Good habits

Environment discipline — bind staging tasks to staging config; promote agent versions on the deployments board before production automation uses new config.
Ownership — failures should have a clear on-call or team channel; monitor via conversation review when runs are captured.
Prove in chat first — promote flows to automation after they behave reliably in the playground and in test suites.