Assigning Tasks to AI Agents
Assigning tasks to an AI agent works the same way as assigning to a human:
- Open the task you want to assign
- Go to the Assignees section
- Click “Add Assignee”
- Search for your AI agent by name
- Select and save
The AI agent now has access to this task and all of its Linked Content documents.
What the Agent Receives
When assigned a task, the AI agent can access:
- Task details — name, description, duration
- SOPs — the policies governing the task
- Work Instructions — especially structured-format WIs with executable steps
- Related content — any linked training or reference material
Staying Current When Documents Change
Linked content doesn’t go stale. When a controlled document linked to the agent’s task is republished, BFF fires a document_changed webhook event to the agent. The agent then re-fetches the latest published content, so it always executes against the current version — no manual re-sync needed.
Linking Content for AI Execution
For the best AI agent performance:
- Create structured Work Instructions — These give the agent explicit steps to follow
- Link SOPs — These provide context and rules
- Set clear expected outputs — Define what success looks like for each step
AI Agent Task Execution
Once configured, the AI agent:
- Pulls its task list via the self-service API
- Reads the structured Work Instructions for each task
- Executes step by step — following the defined actions
- Reports progress back to BFF (“completed step 3 of 5”)
- Logs results — full execution history is visible in the UI
Company admins can monitor every agent’s activity fleet-wide on the Agent Activity page at /dashboard/agent-logs (admin-only).
Beyond Task Execution: Drafting Documents
Agents aren’t limited to executing tasks — they can also draft documents for human review. See Agent Document Authoring for how agents propose drafts that a human reviews and publishes.
Bulk Assignment
For agents handling multiple responsibilities:
- Assign multiple tasks to one agent
- Schedule different tasks at different times using time blocks
- The agent handles task context-switching automatically
Tip: Start with one well-documented task, verify the agent executes correctly, then expand to additional tasks.