Schedules & Tasks

Not every task starts with a conversation. Some things need to happen on a regular schedule without anyone remembering to ask — weekly reports, daily summaries, automated follow-ups, or routine checks. Schedules and tasks let you tell an agent “do this thing at this time, every time” and then forget about it.

How scheduled tasks work

A scheduled task is a set of instructions that an agent executes automatically at a time you define. The agent runs the task, completes whatever actions are needed (searching knowledge bases, sending emails, generating reports), and logs the results. You can review what happened after the fact, and the output is delivered to a channel you choose.

Schedules come in two timing flavours:

• Recurring — Runs on a repeating cadence: every day at 7am, weekdays at 5pm, every 30 minutes, the 1st of each month, and so on. Any cadence you can describe is supported. A recurring schedule can also have an end date (“every day until Friday”).
• One-off — Fires exactly once at a specific time, then automatically disables itself. This is what powers “remind me tomorrow at 3pm” and “check back in an hour”.

Times are interpreted in your timezone. If a scheduled run cannot fire on time (for example, during an outage), it is skipped rather than delivered stale — by default a run more than 10 minutes late is dropped and recorded as skipped.

Common examples for trade businesses:

• “Every Monday at 8am, generate a weekly job summary and email it to the team.”
• “Every weekday at 5pm, send me a list of jobs that were completed today.”
• “On the 1st of each month, check which customer maintenance contracts are due for renewal and draft reminder emails.”
• “Remind me tomorrow at 7am to organise the EWP for the Slacks Creek job.”

Kinds of scheduled task

Behind the scenes, a scheduled task is one of four kinds:

Kind	What happens when it fires
Reminder	The exact message you wrote is delivered verbatim to your chosen channel. No AI processing, so reminders are instant and free.
Instruction	The agent carries out a natural-language instruction (“compile today’s completed jobs and email the summary”) and delivers the result.
Script	A saved script runs in the agent’s sandbox — useful for deterministic data jobs.
Pipeline	Data is moved from a source to a destination, optionally transformed along the way, with no AI involvement.

You do not usually need to think about these — when you ask an agent to schedule something, it picks the right kind. “Remind me” requests become reminders; “do X every morning” becomes an instruction task.

Creating a scheduled task

Schedules are created in chat — just ask the agent that should own the task. Choose an agent whose role matches: a reporting task belongs with your admin agent, a follow-up task with your customer service agent.

1. Open a chat with the agent

   From the Agents page, open a conversation with the agent you want to assign the task to.

2. Describe the task and the timing

   Tell the agent what to do and when, in plain language. Be specific about what output you expect and where it should go. For example:

   Every Monday at 8am, search the job management system for all jobs
   completed in the past 7 days. Summarise each job with the customer name,
   job type, technician, and total cost. Format the summary as a clear list
   and email it to operations@smithplumbing.com.au with the subject line
   "Weekly Job Summary - [date range]".

3. Confirm the schedule

   The agent creates the schedule and confirms the next run time back to you. You can ask it to list its schedules at any time to double-check what is set up.

Try asking your agent

“Schedule a daily job summary every morning at 7am and email it to the team.”

“Remind me tomorrow at 3pm to call the supplier about the pump order.”

“Set up a weekly task to check for overdue invoices every Monday at 9am.”

Tip

Write task instructions the same way you would explain the task to a capable employee. The more specific you are about what to do, where to find the data, and where to send the results, the better the output.

Reminders in team chat

If you set a one-off reminder from inside a team chat room, the reminder posts back into that room when it fires (as the agent) and pings anyone it @-mentions. Ask for the reminder “privately” if you want it delivered to you alone instead.

Managing schedules

An agent’s schedules are listed in the Schedules section of its detail page. Each entry shows the timing in plain language (“Weekdays at 7:00 AM”, or “One-shot” for one-off tasks) and whether it is enabled. From there you can:

• Pause or resume a schedule — Toggle the schedule off to temporarily stop it from running, and back on when you want it again. The task and its history are preserved.
• Delete a schedule — Remove the task entirely.

To change a schedule’s timing or instructions, just ask the agent in chat — “move the weekly summary to 9am” or “also include the technician’s hours in the daily report”.

Viewing execution history

Every time a scheduled task runs, Sprigr Teams logs the execution. The Task History section of the agent’s detail page shows the agent’s recent background runs. Each entry shows:

• Status — Whether the run is queued, running, completed, or failed. Runs that fired too late are recorded as skipped.
• Duration — When the run started and how long it took.
• Usage — How much work the run involved.

For the full step-by-step breakdown of what the agent did during a run — tool calls, searches, and API requests — see the activity log.

Caution

If a task fails repeatedly, check the task instructions for ambiguity. Common issues include referencing a knowledge base that does not exist, trying to send emails without a connected email integration, or instructions that are too vague for the agent to act on. The nightly Insights review also flags enabled schedules whose last run failed.

Tips for effective scheduled tasks

• Start simple. Begin with one or two scheduled tasks and expand once you see the results. A weekly job summary is a great first task.
• Match agent to task. Assign tasks to agents that have the right tools and integrations connected. A reporting task needs access to your job management system; an email follow-up task needs a connected email integration.
• Review the output. Check the task history regularly for the first few weeks to make sure the agent is producing the results you expect. Adjust the instructions if needed.
• Combine with workflows. For complex processes that need approvals or multiple agents, consider using a workflow with a schedule trigger instead of a simple scheduled task.

Next steps

• Agent Settings — Configure the tools and integrations your agent needs to execute scheduled tasks.
• Activity Log — Understand the detailed breakdown of what your agent does during task execution.
• Workflows — For multi-step processes that need approvals, workflows with schedule triggers are a more powerful option.
• Integrations — Connect email, job management, and other services so your scheduled tasks can take real actions.
• Deep Tasks — For assignments too large for a single scheduled run, a task can hand the work to a background specialist.