AI agent examples from one person's projects

AI agent examples tend to be enormous: an agent that runs a company, an agent that books your whole life. The useful ones are smaller. Here are seven from ordinary personal projects, each with what you ask, what the agent does, what it reports, and where it stops to ask you.

How to read these examples

Each example is one ticket, or a handful. The agent works inside one project and reports in the ticket's thread. It uses statuses you will recognise: in progress while it works, blocked when it needs you, done when the work is finished, and deployed only for things confirmed live. All of it happens on a screen. It cannot pick up a paintbrush, and you would not want it to.

1. A website that needs a tidy-up

• You ask: "Check every page for broken links and prices that do not match the price list. One ticket per problem."
• The agent does: goes through the pages and files a ticket for each broken link and each wrong price, with the page it was on.
• It reports: a short summary in the original ticket, with the numbers of the tickets it filed.
• It asks: "Two pages link to a product you no longer sell. Remove the links, or point them at the new product?"

You get a clean list instead of a wall of text, and you can fix the easy ones yourself or hand them back.

2. A freelance client project

• You ask: "Here is the client's feedback email. Turn it into tickets, one per change."
• The agent does: splits the email into separate changes, writes each as a ticket with the client's own words quoted in the body.
• It reports: which points it could not place, such as a comment that might mean two different things.
• It asks: "The client says the logo should be bigger in one place and smaller in another. Which page do they mean?"

The reply to the client stays with you. The agent can draft it, but you send it.

3. Research for a home renovation

• You ask: "Compare these three quotes for the bathroom. What is included, what is missing, and what is priced differently?"
• The agent does: reads the quotes you gave it and builds a side by side comparison in the ticket.
• It reports: the gaps, such as one quote that leaves out removing the old tiles.
• It asks: "None of the quotes mention the ventilation. Should I add a ticket to ask each contractor?"

If it mentions rules or permits, treat that as a question to check, not an answer. The decision and the phone calls are yours.

4. Organising a move

• You ask: "Make a ticket for everything that has to happen before moving day, with the ones that depend on a date first."
• The agent does: files tickets for address changes, notices, utilities and packing, based on what you told it about your situation.
• It reports: which tickets have a deadline and which can wait.
• It asks: "Do you have a date for handing over the keys? Three tickets depend on it."

Once the list exists, you are doing ordinary personal project management, just with a faster start.

5. A newsletter

• You ask: "Go through my notes from this month and draft the next issue. Keep it under the usual length."
• The agent does: picks the items that fit, writes a draft, and puts it in the ticket.
• It reports: which notes it left out and why.
• It asks: "Two items contradict each other on the launch date. Which one is right?"

The agent marks the draft done. It is not deployed until you have read it and pressed send.

6. A small online shop

• You ask: "Find products with missing sizes, missing photos or a description under two lines."
• The agent does: checks the product list and files one ticket per product that needs work.
• It reports: the products it could not check, for example because they are hidden.
• It asks: "Should I draft the missing descriptions, or only list them?"

Prices and discounts are a good example of something that waits for you. An agent that can change them is an agent that can change them wrong.

7. A project that has gone quiet

• You ask: "Read every open ticket in this project. Which are duplicates, which are out of date, and which still matter?"
• The agent does: goes through the list and writes its view in one ticket, grouped into keep, merge and cancel.
• It reports: the few tickets it thinks you should look at first.
• It asks: "Cancel these eight, or would you rather go through them yourself?"

This one pairs well with a plan for managing several projects at once, because quiet projects are where old tickets pile up.

What these AI agent examples have in common

• Small, defined work. One ticket, one job, and a clear idea of what done means.
• A project to report into. The work lands in a thread you can read later, not in a chat that scrolls away.
• Questions as a status. When the agent needs you, the ticket says blocked, with one concrete question.
• Decisions stay with you. Anything that costs money, speaks for you or cannot be undone waits for your answer.

None of this changes what personal project management is. You still own the projects and the choices. You just have one collaborator who can do some of the work in between.

Very Simple Projects is a simple personal project management app where you and your agent share exactly this kind of project: numbered tickets, a thread on every ticket, and six statuses you can add to. Each agent gets one project through its own revocable token, and every status change lands in the thread with its name. It cannot delete comments or see projects you did not give it. There are no team seats. It costs $0.99 a month, and the first month is free.