--- title: "How to Schedule a Recurring AI Agent Job" url: https://www.insulin.dev/blog/how-to-schedule-a-recurring-ai-job/ canonical: https://www.insulin.dev/blog/how-to-schedule-a-recurring-ai-job/ type: Blog description: "How to schedule a recurring AI agent job in Insulin: pick the agent, write the task in plain English, choose the trigger, and review the runs." --- # How to Schedule a Recurring AI Agent Job > Canonical HTML version: https://www.insulin.dev/blog/how-to-schedule-a-recurring-ai-job/ 1. [Home](/) 2. / 3. [Blog](/blog/) 4. / 5. How to Schedule a Recurring AI Agent Job # How to Schedule a Recurring AI Agent Job Turn a chat request into an unattended schedule: pick the agent, write the task in plain English, choose the trigger, review the runs. A step-by-step walkthrough. ![Shirley Guo](/authors/shirley-guo.jpg) Shirley Guo Aug 18, 2026 ![How to Schedule a Recurring AI Agent Job](/images/blog/how-to-schedule-a-recurring-ai-job/hero.png) Explore AI Summary [![](/logos/company/openai.svg)](https://chat.openai.com/?q=Read%20and%20summarize%20https%3A%2F%2Fwww.insulin.dev%2Fblog%2Fhow-to-schedule-a-recurring-ai-job%2F%2C%20then%20cite%20the%20source.%20Focus%20on%20what%20it%20says%20about%20Jobs%2C%20Automation. "Summarize with ChatGPT")[![](/logos/company/anthropic.svg) ](https://claude.ai/new?q=Read%20and%20summarize%20https%3A%2F%2Fwww.insulin.dev%2Fblog%2Fhow-to-schedule-a-recurring-ai-job%2F%2C%20then%20cite%20the%20source.%20Focus%20on%20what%20it%20says%20about%20Jobs%2C%20Automation. "Summarize with Claude")[![](/logos/company/gemini.svg)](https://www.google.com/search?udm=50&aep=11&q=Read%20and%20summarize%20https%3A%2F%2Fwww.insulin.dev%2Fblog%2Fhow-to-schedule-a-recurring-ai-job%2F%2C%20then%20cite%20the%20source.%20Focus%20on%20what%20it%20says%20about%20Jobs%2C%20Automation. "Summarize with Gemini")[](https://www.perplexity.ai/search/new?q=Read%20and%20summarize%20https%3A%2F%2Fwww.insulin.dev%2Fblog%2Fhow-to-schedule-a-recurring-ai-job%2F%2C%20then%20cite%20the%20source.%20Focus%20on%20what%20it%20says%20about%20Jobs%2C%20Automation. "Summarize with Perplexity") Table of Contents - [Step 1: Pick the agent](#step-1-pick-the-agent) - [Step 2: Write the task in plain English](#step-2-write-the-task-in-plain-english) - [Step 3: Choose the trigger](#step-3-choose-the-trigger) - [Step 4: Review the runs](#step-4-review-the-runs) - [What a job inherits from its agent](#what-a-job-inherits-from-its-agent) - [Frequently asked questions](#frequently-asked-questions) - [Takeaways](#takeaways) _Scheduling a recurring AI job means turning a request you would type into chat into an agent that runs it unattended — on a timer, one time, or on a business event. In Insulin the recipe is four decisions: pick the agent, write the task, choose the trigger, review the runs._ * * * There is a moment every team hits with a useful AI request. You have asked the agent the same thing three Mondays running — reconcile these disbursements, check these listings for drift, summarize what changed — and it works every time. The only thing wrong is that _you_ keep having to be the one who remembers to ask. That is the exact point where a request becomes a job. Insulin’s Jobs page puts it plainly: **“If you can describe the task in chat, you can schedule it. Jobs use the same plain-English instructions.”** You are not learning a new tool. You are taking something that already works interactively and moving it onto a schedule. Here is the walkthrough, one decision at a time. * * * ## Step 1: Pick the agent The first decision is _who_ runs the job, not what it does. Every job runs an existing Insulin agent, and the Jobs page is explicit about why that comes first: the step is to **“Choose the specialist that already has the instructions and integration scope for this work.”** This matters more than it looks. A job has no permissions or knowledge of its own — it borrows everything from the agent you name. So the finance agent that already reaches your ledger and knows your reconciliation rules is a better starting point than a generic assistant you would then have to configure. Pick the agent whose day job this already is. If no agent fits yet, that is a signal to build or install one _before_ you schedule anything. Start from [Insulin’s AI agents for marketplace operations](/agents/) or install one from the [AI agent marketplace](/agent-marketplace/), get it working in a normal chat, and come back. The job inherits whatever state the agent is in. * * * ## Step 2: Write the task in plain English The second decision is what the run should do, described exactly the way you would ask in chat. The Jobs page frames the step as **“Describe what the run should do in plain English, the same way you would ask in chat”** — and that is not a simplification, it is how creation actually works. Per the [Suger Insulin Jobs docs](https://doc.suger.io/insulin/jobs/), you create a job by asking the built-in assistant to automate a task (or via the REST API); there is no separate “Create Job” form to fill out. So the task is a sentence, not a configuration screen. A good one is specific about inputs, comparison, and output: > Compare yesterday’s marketplace disbursements with the finance ledger. List every record that does not match and explain why. Notice what that instruction does _not_ say: it does not name a database connection or a file path, because the agent you picked in Step 1 already carries those. You are describing the work, not wiring it up. Write it the way you would explain it to a competent colleague who already has access to the systems — because, functionally, that is who is running it. * * * ## Step 3: Choose the trigger The third decision is _when_ it runs. Insulin gives a job three trigger types: - **One-time** — a single unattended run, for a backfill, an audit, or a one-off sweep. You pick the results up when it finishes. - **Recurring** — the agent on a timer, so the work happens “every morning, every Monday, or at month-end without anyone remembering to ask.” Under the hood this is a cron schedule. - **Event-driven** — a change in the business starts the run, “so the agent reacts when something actually happens instead of waiting for the next tick of the clock.” For a _recurring_ job — the subject of this walkthrough — you choose the schedule and let it go. Daily at 07:00, Mondays at 09:00, the first of the month: the cadence is yours. The whole value of the recurring trigger is that the run “happens whether or not anyone remembers it,” which is the entire reason to move deadline-shaped work off a person’s to-do list. Choosing between the three is its own decision, and getting it wrong is why automation either arrives too late or runs constantly for nothing. If your work is defined by a deadline, schedule it; if it is defined by an event, listen for the event. That trade-off is worth reading before you commit — see [scheduled vs event-driven timing for AI jobs](/blog/scheduled-vs-event-driven-ai-jobs/). * * * ## Step 4: Review the runs The fourth decision is what happens after it fires. The Jobs page describes the loop as **“Read the output, adjust the task, and let the job keep going”** — a job is not fire-and-forget, it is a running process you tune. The first few runs are the ones that earn your trust. Read them the way you would read a new hire’s first reports: is the output right, is it landing where the team already looks, is the task phrased tightly enough that the run does the same thing every time? If a run is off, you edit the task in plain English — the same sentence from Step 2 — and let the next tick pick up the change. Nothing gets re-trained; you are refining an instruction, not rebuilding a pipeline. Once the runs are boring in the good way — same output, same place, every time — you can mostly stop watching. That is the point. The Monday digest arrives before the Monday meeting whether or not you thought about it. * * * ## What a job inherits from its agent A job inherits the agent’s integration scope and any attached knowledge bases. This is the single most important thing to understand before you leave one running, and the Jobs page states it directly: **“A job inherits the agent’s integration scope and any attached knowledge bases.”** Two consequences follow: - **Scope.** A job “runs within the creating user’s scope, so the agent has access to the same resources and integrations that user does,” per the [Jobs docs](https://doc.suger.io/insulin/jobs/). An unattended run reaches exactly the connected systems the agent reaches in an interactive session — no more, no less. That is what makes it safe to leave running: the schedule does not widen access, it just removes the requirement that you be present. - **Grounding.** If the agent has a knowledge base attached, the job reads from those same grounded documents — your runbooks and policies, with cited sources — when it runs at 07:00 with nobody watching. So an unattended answer is as auditable as an interactive one. If you have not grounded the agent yet, do it first: attach an [AI knowledge base](/knowledge-bases/) to the agent, then schedule the job on top of it. The design principle is that jobs “do not have their own permissions or their own knowledge. They inherit everything from the agent you pick, which is what makes them safe to leave running.” Get the agent right and the job is right by construction. * * * ## Frequently asked questions **How do I schedule a recurring AI agent job in Insulin?** Pick the agent that already does the work, write the task in plain English the way you would ask in chat, choose the recurring trigger and its schedule, then review the first runs and adjust. Jobs are created conversationally, not through a form. **Do I need to configure integrations for a scheduled job?** No. A job inherits the agent’s integration scope and any attached knowledge bases. It runs within the creating user’s scope, so an unattended run reaches the same connected systems and grounded documents the agent reaches in an interactive session. **How do I write the task for a job?** Describe what the run should do in plain English, the same way you would ask in chat. Be specific about inputs, comparison, and output. You do not name connections or file paths — the agent you picked already carries those. **What trigger should I choose for a recurring job?** Use the recurring trigger for deadline-shaped work — daily, weekly, or month-end. Under the hood it is a cron schedule. Choose one-time for a backfill or audit, and event-driven when a business change should start the run. **Can I change a job after it is scheduled?** Yes. Reviewing runs is part of the loop: read the output, adjust the task in plain English, and let the job keep going. The next run picks up the change with no re-training or pipeline rebuild. * * * ## Takeaways - Scheduling a job is four decisions: pick the agent, write the task, choose the trigger, review the runs. - Pick the agent first — a job borrows its permissions, integrations, and knowledge, so the specialist that already does the work is the right start. - Write the task in plain English. Jobs are created conversationally, not through a form: if you can describe it in chat, you can schedule it. - Choose the trigger by the shape of the work — recurring for deadlines, one-time for backfills, event-driven for reactions. - A job inherits the agent’s integration scope and attached knowledge bases, which is what makes an unattended run as scoped and auditable as an interactive one. Ready to move a repeating request onto a schedule? Start by scoping the right [Insulin agent](/agents/), ground it with an [AI knowledge base](/knowledge-bases/), then see how [scheduled AI jobs](/jobs/) run it unattended — once, on a timer, or on a business event. ## Sources Primary sources for the platform rules cited above. Last verified August 18, 2026. Cloud providers change fees, eligibility, and program terms without notice — check the source before relying on a figure. - [Suger Insulin docs: Jobs](https://doc.suger.io/insulin/jobs/) — Jobs are created by asking the built-in assistant in chat or via the REST API; the manual, cron and push trigger kinds; and that a job runs within the creating user's scope, reaching the same resources and integrations that user does. ### Stay Updated Get the latest Cloud GTM insights, product updates, and marketplace strategies delivered to your inbox.