Blog
18 AI agent examples from everyday team work
Eighteen AI agent examples, grouped by team. Each is a real one-line ask and the result that comes back, with the steps that still wait for a yes.
Most lists of AI agent examples describe categories. "A sales agent." "A support agent." That tells you what the vendor wants to sell and very little about what you would type on a Tuesday morning.
This list goes the other way. Each example is one sentence someone types into a team chat and the thing that comes back. They are grouped by team because that is how the work arrives. Every one of them is a job Cedric does today, so where a result is described, it is the result he produces.
A working definition first. An AI agent reads your tools, decides what the task needs, and does the work across them, instead of answering from memory. The good ones also know which steps to stop at. In every example below, anything that sends, changes a record or spends money waits for a person to say yes.
- An agent earns the name when it reads your real tools and does the steps, not when it writes a nice paragraph.
- The best asks are short. "Chase the overdue invoices" beats a page of instructions.
- Reading and drafting run on their own. Sending, changing records and spending wait for a yes.
- Many of these jobs recur. Say "every Monday at 9" once and the agent keeps doing them.
- Start with the job your team complains about most, then add the next one.
Sales
Qualify a new lead. Type "who is this new lead?" and Cedric sweeps the mailboxes, the contact form and the CRM, looks the company up on its own public pages, and comes back with who wrote in, what they want and whether they fit, with the reason. The reply is drafted. It goes out on your yes. See qualify a lead.
Turn a call into CRM updates. "Put this call in HubSpot." Cedric was in the call, so he proposes the summary, the new stage, the next step with a date and the contact he just met. One approval writes it, with the call attached as the source. More on meeting to CRM.
Do the follow-through after a deal call. "The Steiner AG call is done. Do the follow-up." Back comes the opportunity updated, the email drafted, the task and the reminder set, each one waiting for your yes.
Marketing
Write the campaign from a launch note. "Turn this launch note into the campaign." You get the email draft, the posts and a schedule for the week. On Friday he reports what ran and what to change next week, read from the real numbers. See the AI marketing agent.
Track competitors. "What did our competitors do this week?" Cedric checks changelogs, careers pages and press pages, posts the moves that matter with a source on each, and keeps a tracker page the team can open. Say "every Monday at 8" once and it runs weekly.
Check the ad account. "Check our Google Ads for last week." Spend, conversions, the search terms that wasted money, and the changes he would make, such as negative keywords or an ad group to pause. Nothing in the account changes without a yes. See the Google Ads agent.
Operations and finance
Chase overdue invoices. "Chase the overdue invoices." Cedric reads what is unpaid from live billing and drafts one reminder per invoice in your voice, with the number and the amount in it. Billing stays read-only. The batch goes out on one yes, and a client's "paid on Friday" reply drops that invoice from next week's list. Details on invoice follow-up.
Onboard a client. "Onboard Steiner AG." This one runs for days. He books the kickoff, sends a short intake form, collects the replies, nudges whoever has not answered and sets up the folder. The welcome email waits as a preview until you approve it.
Collect everyone's weekly report. "Get everyone's reports by Friday." Cedric asks each person in a direct message, counts each reply as it arrives, and posts a roll-up that says who is done and who is still open.
One example, in full
The invoice chase, as it lands in the channel.
Grace Adeyemi9:14 AM
@Cedric chase the overdue invoices.
CedricAPP9:15 AM
Read from Stripe. The oldest is Keller Bau, invoice 2026-118, CHF 3,150, due 24 July. I drafted one reminder per overdue invoice, each in your voice with its number and amount. Nothing goes out until you say yes.
Founders
Catch up after time away. "Catch me up, I was out all week." One briefing comes back in the order you need it. First what is waiting on you, then the decisions made, then the rest, each line with the channel or thread it came from. See catch me up.
Plan the day. "What should I do today?" Cedric reads the calendar, the mail, the channels and the board. He answers with what is fixed, what to do in what order and why, and what only you can unblock. He also says what he could not see, which is the part that makes the rest trustworthy.
Get the weekly numbers. "Send me the numbers every Monday at 9." Billing, the CRM and the tracker in one note, each figure with its source, posted on time. Ask "where did the dip come from?" in the thread and he answers with the invoices behind it.
Engineering
Say what shipped. "What did we ship last week?" The merged pull requests, the notable ones named, and anything merged but not yet deployed. It works in a meeting too, out loud, when someone says his name.
Find why the build broke. "The checkout PR is failing CI, find the cause and open a fix PR with a regression test." The detail in that sentence is what gets you a fix instead of a guess. Merging stays gated.
Use the internal tool. Once your deploy server is connected, "roll us back to yesterday's release" gets a prepared rollback that keeps the current release for a fast return. It changes production, so it waits for a yes.
Meetings
Take the notes. "Join my 2pm and take the notes." He joins as a visible guest and posts the decisions, the open items and the action items with owners when the call ends.
Prep the call. "Prep me for my 2pm." An hour before, a brief lands with who you are meeting, what changed since you last spoke, what to ask, and what to keep out of the room. See meeting prep.
Send the follow-ups. "Send the follow-ups from this call." The recap to the other side as a draft, the action items filed in your tracker with owners and dates, and a nudge when an owner goes quiet. More on meeting follow-ups.
What these have in common
Read the eighteen again and a pattern shows. The ask is one line. The agent does the reading and the drafting without asking, because none of that leaves the team. The moment a step would reach a client, change a record or move money, it stops and shows you exactly what it is about to do.
The other pattern is repetition. About a third of these jobs recur on a schedule. Once a job is right by hand, one sentence puts it on a schedule, and it keeps running until you pause it. The agent gallery has each of them as a ready-made starting point.
What still needs a person
Every example above ends with a decision that stays human. Whether the lead is worth a call. Whether the reminder is too firm for this client. Whether yesterday's release is the one to roll back to. The agent gets you to that decision with the facts in front of you. It does not make it for you.
An agent also only sees what you connect. If the contact form feeds a mailbox Cedric cannot read, he says so. He does not guess.
FAQ
Typing "chase the overdue invoices" in the team chat, and getting back one drafted reminder per unpaid invoice, read from live billing, that go out when you approve them.
A chatbot answers from what it already knows. An agent reads your tools, does the steps across them, and comes back with finished work, stopping for approval before anything leaves the team.
The job people complain about most. For many teams that is the weekly numbers, the invoice chase or meeting follow-ups, because they recur and nobody enjoys them.
Cedric does every job on this list from the same chat, with one memory for the team. You do not set up a separate agent per task. You ask for the next job.
Cedric does not. Drafting is free, and anything that sends, changes a record or spends waits for a yes. See how approvals work.
100,000 free credits on sign-up, no card. Plans from $19 a month for the whole team.
