Send reports on a schedule with the report automation
The report automation is an AI agent that acts on your thunk's saved reports. You tell it, in plain language, what to do with them — "every Monday at 8am, email the Overdue Items report to [email protected] as a spreadsheet" — turn on a schedule, and it runs on its own. You can also ask it for something right now in its chat, without waiting for the schedule.
Because it is an AI agent driven by your instructions, not a fixed list of delivery options, it can do more than forward a file. It can:
Email a link to a report, so recipients see the live results.
Export a report to a CSV or Excel file and attach it to an email.
Write a short summary of what a report shows and send that.
Combine several reports in one message.
Decide based on the date or the data — for example, only send a report on the last business day of the month, or only mention a report when it has results.
Each thunk has one report automation. It covers every saved report on the thunk and reaches each one by its name.
Concepts
How it runs
The report automation runs in two ways, and both follow the same instructions:
On a schedule. At each scheduled time, the agent wakes up, reads its instructions, looks at the thunk's saved reports, and does what the instructions say for that moment.
When you ask. Type a request into its chat — "send the Throughput report to me now" — and it runs immediately.
Scheduled runs never read the requests you typed into the chat. A one-off request such as "email this to Bob" is done once, when you ask, and is not repeated on the next scheduled run.
The automation never finishes. After each run it waits for the next scheduled time or the next request in its chat.
Links and files
The automation can deliver a report in two forms, and your instructions can ask for either or both:
Deliver as | What the recipient gets | Use it when |
A link | The report's shareable page, showing its current results each time it is opened, with an AI-written summary | The reader will look at the latest data or drill in |
A file (CSV or Excel) | A spreadsheet of the report's results at the moment it was sent, with the same columns the report shows | The reader wants the data itself, to keep or work with offline |
When it sends a file, the agent exports the report and attaches it without reading the results into its conversation, so even a large report can be delivered.
If a report's filter uses a placeholder such as {{owner}}, the agent can send each person a link that fills in their own value. See Share a report in How to save a report from your work items.
Whose access it uses
The automation works with the thunk owner's access, on scheduled runs and on chat requests alike, whoever set it up or typed the request. A report it sends contains what the owner can see, so recipients can receive work items they could not open themselves. Choose recipients with that in mind.
Details
Open the report automation
Open your thunk.
In the left navigation, open Monitor.
Open the Reporting tab.
The report automation is the side panel beside your list of reports. It has two tabs:
Execution — the automation's chat. Every run, scheduled or requested, appears here, and you type requests here.
Settings — its instructions, the tools it can use, and its schedule.
Only a thunk's owner and admins can open Monitor. The automation is set up the first time an owner or admin opens the Reporting tab.
Write its instructions
On the Settings tab, write what you want done under Directions. Name each report exactly as it is saved, and say who should receive what, in what form, and when. For example:
Every Monday, email the Overdue Items report to [email protected] as an Excel file, with a two-sentence summary of what changed.
On the last day of each month, email a link to the Monthly Throughput report to [email protected].
Tips:
Use the report's saved name. If an instruction names a report that does not exist, the agent says so in its chat and lists the reports that do exist, instead of guessing.
Renaming a report breaks instructions that use its old name. Update the directions when you rename or replace a report.
One schedule, many cadences. The automation has a single schedule. To send different reports on different days, schedule it daily and say in the directions which report goes out on which day.
Let it send email
Sending email is a tool the agent must be allowed to use. On the Settings tab, open AI Tools and make sure the Communications library is on, with sending a new email turned on — it starts turned off. See AI Tools.
Set the schedule
On the Settings tab, open Scheduled Work Settings:
Set Allow scheduled work? to Run on a schedule.
Choose a Schedule — Daily at specific time, Every hour, Monthly on a specific day, or Yearly on specific months — and its time and Timezone.
Changes save as you make them. The automation starts with no schedule, so it never runs on its own until you turn one on; until then it only answers requests in its chat. The schedule options are the same as for a workflow step — see Scheduled work on workflow steps. The most frequent schedule is once an hour.
Run only when there is something to report
Each scheduled run is a full AI agent run, and it uses credits even when there is nothing to send. You can add a Query Filter to the schedule so a scheduled run starts only when at least one of the thunk's work items matches a condition — for example, Status is Overdue. When nothing matches, the run is skipped and costs nothing.
The Query Filter lists only the work-item properties added under Input properties on the Settings tab. To filter on a property:
On the Settings tab, open Input properties and add the property, for example Status.
In Scheduled Work Settings, open Query Filter and add your condition.
The condition looks at all of the thunk's work items, not at one report. To send a report only when that report has results, say so in the directions instead — the agent checks when it runs.
Use separate instructions for scheduled runs
By default, scheduled runs and chat requests use the same Directions. To give scheduled runs their own instructions, write them under Custom AI instructions for scheduled runs in Scheduled Work Settings. Those instructions replace Directions on scheduled runs only; chat requests still use Directions. Leave the box blank to use Directions for both.
Test it
Ask in chat. On the Execution tab, type a request — "send me the Overdue Items report now" — to check that the agent finds the report and can send email.
Test the schedule. In Scheduled Work Settings, choose Test to fire one scheduled run now. You can set Simulate firing at to check date-based instructions, such as "on the last day of the month", without waiting. When a Query Filter is set, choose whether the test should respect it.
Each run's result appears on the Execution tab.
Things to know
Copying a thunk does not copy its report automation. The copy starts with an empty automation: write its directions and set its schedule again.
Recipients of a file get a snapshot. A spreadsheet shows the results at the moment it was sent. Send a link when recipients should always see current results.