BERD TASK · 0017 · AUTOMATE AND DELEGATE

Build an automation by chatting¶

This page is not finished

Automations run unattended, on a schedule, against real systems and a real provider bill; no source says what a bad run does or how you notice, and nothing here has been run hands-on. What is below is the shape of the task and the facts the sources support — not a walkthrough anyone has followed.

A saved, re-runnable task with a schedule and a run history, built by describing it rather than filling in a form.

Not in the public build

At v0.6.4 the public release compiles automations off (VITE_AUTOMATIONS=0 in the release workflow), so the Automations surface is not in the sidebar; Block's internal distribution has it. This page describes what exists in the source for the day a distribution you run enables it. If you need work done on its own today, use the bundled berd-monitor and berd-orchestrator skills in Run long work outside the turn. Last checked: Berd v0.6.4, 2026-09-11.

Steps at a glance

Describe what you want in the builder
Revise the draft and approve it
Set the schedule and time zone
Run it once and read the history
Pin its latest output to Home
Diagnose a failure by phase
Check the work happened, not that the run finished

1. Describe what you want in the builder¶

Describe the first time you opened the builder, and what you typed.

Add automation opens a builder chat that asks you to describe what to automate; it runs on a provider and model like any chat.
Morgan, Berd's design lead, on Block's On The Block (September 2026): asked her agent for a code review every 20 minutes; saves about two hours a day.

[SCREENSHOT: T17-01 — the builder with a draft]

2. Revise the draft and approve it¶

Describe the draft, what you changed, and what approving did.

The draft previews beside the chat; an invalid draft is rejected.
Approval makes it a tile; unsaved changes prompt save or discard.

3. Set the schedule and time zone¶

Describe the schedule editor and what you chose.

Presets hourly, daily, weekdays, weekly, none; Custom takes a cron expression; time zone per automation.
A Slack notifications field exists in the strings; not verified in the public build.

[SCREENSHOT: T17-02 — the schedule editor]

4. Run it once and read the history¶

Describe the first run: where you looked, what appeared, how long it took.

Run now starts a run; History lists runs with output and a session id; Details shows status, last successful run, tools, connections.

[SCREENSHOT: T17-03 — run history]

5. Pin its latest output to Home¶

Describe the pinned widget after a run.

Right-click the Home canvas to pin an automation's latest output.

[SCREENSHOT: T17-04 — pinned output on Home]

6. Diagnose a failure by phase¶

Describe the first failure and which phase it was in.

Drafting fails like a chat: a provider problem.
Approving or saving fails like a product problem, such as an unsupported tile type.

7. Check the work happened, not that the run finished¶

Describe how you verified the effect of a run.

A run marked complete says nothing about whether anything happened — the rule Buzz or Berd applies to both.

Last checked: Berd v0.6.4, 2026-09-11 — read from source and release notes, not yet tested in the app.

Learn more

Reference

Moment: apply
Takes: untimed
Steps: 7
Verified: Berd v0.6.4 · 2026-09-11 · source-read