14 AI Agent Examples: Prompted, Scheduled, and Triggered
Most AI-agent examples begin with a prompt. That can still be agentic work—the system may plan and execute several steps—but it does not tell you whether the job can start later without you.
A more useful distinction is what starts the run: you, a clock, or an event in another system. The products now overlap across those categories, so plan, integrations, and approval settings matter as much as the model. Below are 14 concrete examples and the conditions each one needs.
The three kinds, and why the difference is the whole story
ChatGPT Scheduled Tasks and Claude Cowork both run unattended on a timer. Gemini Spark goes further with Gmail-filter and topic monitors for eligible personal paid accounts. ChatGPT Workspace Agents can be started through an API, but another system still has to notice the event and make that call.
Carly supports a broader built-in trigger set: incoming email, calendar changes and offsets, webhooks, payment and CRM events, page or RSS changes, and new spreadsheet rows. That makes “event-triggered” a spectrum rather than a yes-or-no label. Ask which events a product can hear, which apps it can act in, and whether a write step can finish without pausing for approval.
Which tool fits which trigger?
ChatGPT and Claude Cowork are strong for prompted and scheduled work. Gemini Spark now adds Gmail and topic monitors, although it is limited to eligible personal accounts, supported regions, and paid plans; Google also warns that some tasks may require confirmation and schedule times are approximate.
Carly is the broader no-code option when a workflow must begin from email, a calendar event, a webhook, a payment or CRM change, a page update, or a new spreadsheet row and then continue across several apps. Its workflow canvas includes waits and explicit human-in-the-loop steps, so approval can be added where the action warrants it.
Carly has roughly 260 native connectors and can call other public APIs with your own key. Up to 10 Zapier-style workflows run free with no cap on runs; AI agents start at $35/month.
Prompted examples: you are the trigger
These are broadly feasible when the assistant can access the source material and the relevant connected apps. Connector availability still decides whether an example is a real cross-tool workflow or a copy-and-paste exercise, and each one requires you to start the run.
- Summarize a long thread and draft the reply. Point the assistant at a 40-message thread, get the state of play and a draft response you review.
- Turn a messy meeting into decisions and owners. Feed it a transcript, get a decision log rather than a wall of notes.
- Pull a report across two tools. Where both services are available as connected apps, ask for last month’s closed deals from the CRM against actual revenue from your payment processor, in one table.
- Clean up and enrich a list. Deduplicate a spreadsheet of contacts, fix formatting, fill in missing company fields.
- Draft a first-pass document from source material. A proposal from a discovery-call transcript, a spec from a scattered set of notes.
Useful, all of them. Also: every one waits for you to open a window and ask.
Scheduled examples: the clock is the trigger
A real step up. Scheduled work is now available across ChatGPT, Claude Cowork, Gemini, and Carly, generally on paid plans. What differs is which connected apps a scheduled run can use and whether a write action can proceed unattended.
- The Monday numbers report. Every Monday at 8am, pull last week’s figures from analytics and billing and prepare the summary. Depending on the mail connector and approval settings, the system may send it, save a draft, or notify you that it is ready.
- A daily digest of what changed. Each morning, check the handful of things you care about and report only the deltas.
- A recurring housekeeping pass. Weekly, archive stale items, re-tag what was miscategorized, and flag what went quiet—where the connected app exposes those write actions.
- A morning meeting-prep batch. Each weekday, assemble briefs for that day’s calls from the calendar, recent email, and account history.
The clock can still be the wrong trigger. An hourly job may learn about an event up to an hour late, while an event monitor can start as soon as the qualifying change arrives.
Event-triggered examples: something happening is the trigger
Event-triggered workflows matter most when response time matters, because the event itself starts the run instead of waiting for the next polling interval.
- A client emails asking to meet. The agent reads the request, checks live availability, proposes workable times, replies, and books the selected slot. See an agent that books appointments from email.
- A new lead comes in. It is qualified, logged in the CRM, assigned, and sent a personal first reply according to the configured workflow. See an agent that follows up with leads.
- A meeting recap or transcript lands. The notes are cleaned up, the CRM record is updated, the follow-up is prepared or sent, and the next step is scheduled. See an agent that updates your CRM after meetings.
- The billing system marks an invoice overdue. The configured reminder goes out, the activity is logged, and the account owner is notified. See an agent that sends invoices and reminders.
- Mail lands in a busy inbox. It gets read, sorted, and the routine ones answered, so what is left is the mail that actually needs you. See an agent that triages your inbox.
The same pattern, by job
The examples above are generic on purpose. What they look like in practice depends on what lands in your inbox all day:
Executive assistants · Consultants · Recruiters · Real estate · Lawyers · Financial advisors · Agency owners · SDRs · Account executives · Product managers · Customer success · Ops · HR and people ops · Marketing managers · Podcasters · Course creators · Newsletter writers · Ecommerce sellers · Airbnb hosts · Wedding planners · Personal trainers · Salon owners · Tutors · Nonprofit directors
What these workflows require
- A defined start condition. A prompt, a schedule, or a specific event.
- The right app permissions. Reading an email and sending one are separate capabilities.
- A completion condition. The workflow should say what counts as finished and when to stop.
- Intentional approval gates. Routine actions can run unattended; sensitive sends, deletes, or financial actions may deserve a review step.
Building these yourself
Prompted examples need an assistant with access to the source material. Scheduled examples need a qualifying plan, connectors, and permissions. Event-triggered examples need a native monitor, a platform such as Carly with app and webhook triggers, or an externally wired API call. For sensitive actions, add a review or resume step rather than assuming every run should finish silently.
FAQ
What is an AI agent, in one sentence? Software that takes a goal rather than a command, decides the steps itself, and carries them out across your tools. What is an AI agent covers the definition and where the term gets stretched.
What is the difference between an AI agent example and a use case? In practice, nothing. “Use case” tends to describe the job in the abstract (“inbox triage”), while “example” describes a specific run of it (“a client emails asking to meet, and here is what happens next”). The examples above are written the second way because the details are where these things succeed or fail.
Can AI agents work while I am asleep? Yes, if the product supports remote scheduled or event-triggered runs. A closed laptop is no longer the dividing line; the real questions are what starts the run and whether required write actions can proceed under the configured permissions.
Which of these examples can I actually build without code? Every example on this page can be assembled without writing code, but not in every assistant or on every account. Prompted examples are broadly available. Scheduled examples require a supported plan, connectors, and permissions. Event-triggered examples require either native monitors, a platform such as Carly with app and webhook triggers, or an externally wired API trigger. Sending and financial examples may also require approval depending on the product and workspace policy.
Ready to automate your busywork?
Carly schedules, researches, and briefs you—so you can focus on what matters.
See what people say
"Before Carly, I relied on a Calendly link, but the whole process felt impersonal and not very professional. Carly changed that by handling all the back-and-forth, so I'm no longer stuck in endless email threads trying to line up schedules.
Now Carly reaches out to candidates, shares my real-time availability, lets them pick a slot, then sends a Zoom link and drops it straight into my calendar. She sends reminders to both of us before each call, which has significantly reduced no-shows and last-minute confusion.
On top of scheduling, Carly acts like a full executive assistant, sending me my schedule the night before so I can prepare for each call. It reminds me of the old x.ai assistant, but Carly is noticeably smarter, faster, and better suited to my healthcare recruitment business."
