What is the short answer?
Pick a boring task from a trade in your own work history. Run the task by hand for a real owner first, then build the automation for the repeat parts. Leave the logo for later.
Pick a trade from your own work history. Pick a boring task inside that trade. Run the task by hand for a real owner. Build the automation for the repeat parts after that. Leave the logo for later. Leave the website for later. Leave the tool stack for later.
Is that the exciting version? Is it the version with a real shop on the other end? Start with that version.
What should you decide before anything else?
Read the definition post for the basics of the work.
Decide on the trade. Look at your own work history for it. Where did you work? Which shops do you know from the inside? Which owners would answer a text from you?
Read the agency versus product post for that choice. Decide whether you want this at all. Do you like talking to owners? Do you like fixing messy processes? Can you handle a quiet inbox for a while? Read the post on whether to start an agency if the answer is a shrug. Read the post on a solo AI consulting business for the advice path.
Write the trade down in a short line. Keep it narrow. Pick the roofers in your county over the home services in general. Pick the dental offices in your city over the healthcare in general.
How do you pick the task to automate?
Look for a task with heavy repetition. Look for a dull task. Look for a task with a low risk from a wrong answer.
Ask the owners about their workload. What eats the evenings? What slips through the cracks in a rush? What do they hand to a person with a sigh?
Look for the boring answers. Is it the missed phone calls? Is it the follow up after a quote? Is it the reminder before an appointment? Is it the reply to a review?
Skip the tasks with a big downside. Skip the medical advice. Skip the questions for a lawyer. Skip the tax answers.
Read the AI automation business ideas for examples of narrow tasks. Run your favorite through the scorer. Cut the idea with weak spots before you build anything.
How do you test the idea by hand?
Run the task by hand for a real owner. Use a spreadsheet. Use your own inbox. Use your own phone.
Write down each step along the way. Where does the information come from? Where does it go? Which steps feel like copy and paste? Which steps need a judgment call?
Keep the scope tiny. Set a clear finish line with the owner. Check the result together at the end.
Ask the owner a plain question after that. Did it help? Would the shop keep it going? Go with the answer, not with your hopes.
When should you set up the tools?
Set up the tools after the manual run, not before. Why automate a process with no demand? Why learn a platform for a task you have not done yet?
Look at your notes from the manual run. Turn the copy-paste steps into automation. Keep a person on each judgment call. Add a check before anything goes out to a customer of the shop.
Keep the stack small. Use the tools from your current toolbox where possible. Add a new tool only for a real gap. Read the post on AI automation vs AI agents before you pick the tools.
Write a short how-to for each automation. Note the location of each automation. Note the off switch for each automation. Keep it in a shared folder with the owner.
How do you land the early clients?
Start with the owner from the manual run. Ask for an intro to a peer in the trade. Ask whether you can mention the shop by name.
Send a note to your contacts in the trade after that. Send each note by hand, not as a blast. Name the task in the note. Ask about the current process at the shop.
Read the post on finding clients for an AI automation agency for the full outreach routine. Read the post on selling AI agents to small businesses for the demo and the pilot.
How do you decide what to ask for?
Look at the result from the manual run. How much time did it save the owner? Did it keep calls or leads from slipping? Use that as the starting point of the talk.
Keep it simple. Write a short scope. Name the task. Name the finish line. Name the check at the end.
Read the post on agency quotes before an email with a number in it.
How do you keep the agency small?
Keep to a narrow trade for a while. Keep to a narrow task for a while. Say no to each custom project outside the trade and the task.
Write the scope for each client. List the task, the inputs and the deadline. Point to it when the requests start to drift.
Keep a simple log. Write the shop, the task, the result and the intro. Look at the log before you pick the next owner to contact.
Frequently asked questions
Do you need to know how to code?
Can you follow a written process and test it with care? Start with the tools at your level. Pick up each new skill along the way.
Do you need a website before the early clients?
Do your early clients come from intros? Use a plain page with your name, the trade and the task. Add a real result from a pilot later.
Should you register a company at the start?
Ask a local accountant about it. Ask about insurance in that conversation. Get the answer before the client work starts.
How narrow should the niche be?
Can you name the trade and the task in a short line? Keep at it until the answer fits a short line. Change it later if the work demands it.
What should you do next?
Try the scorer on your niche idea tonight. Try the generator if you are stuck on the starting point. Use the idea board for a narrow task. Use the newsletter form below this post.