Skip to content
Honen Academy

Docs

Automations

Give an Agent a repeatable task. An automation supplies the instructions and trigger; the selected Agent supplies its model, tools, and permissions.

The example uses Northstar Onboarding Guide to summarize welcome-call preparation and cite the approved playbook. Start with a manual task, review its answer, then choose the recurring schedule that fits your team.

1. Start from the Agent

Open your Agent's Overview and choose Automations → Create new. The Agent is preselected. You can also open Agents → Automations → New automation or choose a starter such as Knowledge base drift.

2. Write one clear task

Give the automation a name. In Instructions, describe the input, work, output, and limits. Use @ to reference a Knowledge Base document or folder the Agent can access.

Read Welcome Call Playbook in Support Handbook. Summarize the four preparation steps for a customer welcome call and link the source. Keep this task read-only. Do not change content or contact people.

Select Northstar Onboarding Guide in Agent. For this example, its Knowledge Base access is Read, limited to Support Handbook; its other content and management permissions and all channels are off. See Create a knowledge Agent to set up the example Agent.

3. Choose a trigger

For a task you will start yourself, select Manual. The editor explains that it runs only when you start it and displays Run now. Keep Active off while preparing the task.

Manual automation with a selected Agent and read-only instructions

For recurring work, select Scheduled, then choose Frequency, Time, and Time zone. Review Next runs to confirm the schedule is what you intend.

Schedule configuration with Active still off

Webhook is the third trigger option. Its editor says to save the automation to obtain its URL and signing secret. This is an incoming automation trigger, separate from outbound webhooks under Developers.

4. Save the automation

Choose Create. The saved automation page gives you its settings, test action, and run history.

The saved page has Settings and Runs tabs and a Save button. Confirm the Agent, instructions, Manual trigger, and Active switch are still correct.

Saved manual automation with Active off and the read-only task

5. Test and check the answer

Review the instructions and Agent access, then choose Test once. Test saves the current form, executes the Agent’s tools, and opens the run. Automatic result notifications are skipped for test runs. This example only reads the playbook.

Wait for Succeeded, then read the result. The example returns research, workspace preparation, attendee confirmation, and blocker checks. Open the cited Welcome Call Playbook and compare the answer with its preparation section.

Succeeded test with four preparation steps and a Welcome Call Playbook citation

Open the named document in Knowledge Base and compare the answer with the source. Expand Worked for… to inspect the recorded activity. Check that it searched for and read the intended playbook.

6. Find the test again

Choose Back, open Runs, and check Show test runs. Select the successful test to reopen its transcript. Tests are hidden from this list until you select that option.

Run history showing the successful test

Return to Settings to confirm Active is still off. Start a manual automation with Run now whenever you need its result. Activate a recurring task only after reviewing its output and intended effects.

Create or inspect an automation through Honen Agent

The Workspace Assistant can configure an Agent and its automation when you have integration-management permission. Name the Agent, source, task, trigger, time zone, and whether the schedule should be active. For example:

Create a paused weekly automation for Northstar Onboarding Guide. Every Monday at 9 AM in America/Los_Angeles, summarize changes to Support Handbook and cite the changed documents.

Check the saved Agent, instructions, Next runs, time zone, and Active switch. A draft Agent and a paused automation do not run scheduled work.

For a failed task, ask the assistant to inspect its settings, recent runs, and recorded failure. A test executes real tools. Request it only when its actions are appropriate. Pausing stops future automatic triggers; it does not cancel a run already in progress.

Keep recurring work useful

Review Runs to see the task’s status, summary, and conversation. Adjust instructions as your process changes, and turn Active off to pause scheduled triggering.

Choose Knowledge base drift for a report of changed sources and affected courses, Nudge overdue learners for a follow-up task, or Sales calls to courses for course-building work. Adapt the starter instructions and give the selected Agent the required access.

For course maintenance, select the relevant sources and grant permission to propose changes on the intended courses. Ask the Agent to prepare a proposal for the course editor or catalog’s publishing review queue. Your reviewers inspect and approve the changes through the normal publishing workflow.

For recorded execution details, use Inspect Agent logs and analytics. Keep the Agent access policy limited to the resources and actions the task needs; test instructions do not disable write tools.