Should You Run the AI Agents Yourself? Honest Answer: Probably Not.

Can I use AI to do my own marketing? It's a fair question, and since I sell the alternative, you deserve a straight answer from me: yes, you can. The tools are good enough now that a motivated owner can run AI agents on their own Google Ads, social posts and website.

Whether you should is a different question. For most owners I talk to, the answer is probably not. Here's why, and here's when I'd tell you to go for it anyway.

Yes, you can use AI to do your own marketing

Start with what's true. AI agents can now operate a browser the way a person does. Anthropic's Claude in Chrome extension can read, click, navigate sites and fill out forms, and it's on paid Claude plans as of this writing. The cloud browser in ChatGPT Work, also on paid plans as of this writing, can do similar work on signed-in sites, and it asks you to sign in yourself through a secure form instead of handing your password to the model.

So the access exists, and it isn't expensive. A paid plan and a free afternoon are enough to get started. Nothing about this is reserved for marketers.

When doing it yourself makes sense

I'd tell an owner to run the agents themselves in these situations:

• You genuinely enjoy it. Some owners like tinkering with this stuff on a Sunday night. If it's a hobby, it isn't really a cost.
• Your needs are simple. One location, a Google Business Profile, a few posts a month, no ads. That's manageable.
• You have a slow season. A lawn care company in January has hours a roofer in May doesn't.
• Cash is truly tight. If you can't afford any outside help yet, your time is the budget. Spend it on the few things that bring in calls.
• You've done marketing before. If you already know what a good ad and a broken conversion tag look like, you'll catch the agent's mistakes.

If two or three of those fit, try it. Start small and see how it goes.

The real cost is your hour

Here's the part owners skip. The AI subscription is cheap. Your hour isn't.

Say you run a made-up six-person electrical contractor, Clayton Road Electric. When you're quoting jobs, walking sites or landing a big commercial customer, your hour is worth far more than whatever you'd pay someone to do marketing. Every hour spent learning to brief an agent, watching it build a Google Ads campaign and fixing what it got wrong is an hour you didn't spend on work only you can do.

Here's roughly what the first month looks like for an owner starting from scratch.

Illustrative. About 20 hours in month one for a beginner, dropping after that if it sticks. Your numbers will vary.

Twenty hours is half a work week. Multiply it by what your hour is actually worth, and that's the real price of doing it yourself.

The learning curve isn't the buttons

Most owners think the learning curve is figuring out which buttons to press. It isn't. The tools are easy to start. The hard part is knowing what to ask for.

An agent does what you tell it, not what you meant. If you don't know that keyword match types in Google Ads behave very differently, you won't tell it which one to use. If you don't know your contact form needs conversion tracking, it won't occur to you to ask. The agent will build exactly what you described, and it will look finished.

That's why briefing matters more than prompting. The AI needs to know your business before it does anything useful, which I walk through in stop prompting, start briefing.

Setup and safety take real time

Before an agent touches anything, there's setup most people don't think about:

• Accounts in your name. Google Ads, Analytics, Business Profile, Meta and your website. You should be the owner of each one, and know how to sign in.
• A separate browser profile. Anthropic's safety guide for Claude in Chrome recommends using a browser profile that has no access to your bank or other sensitive accounts. Your marketing tools go there. Your bank doesn't.
• Approval settings. Decide what the agent can do on its own and what it must ask about. Anything that publishes, sends, spends or deletes should wait for you.
• Logins and CAPTCHAs. Agents get stuck at sign-in screens and security checks. That same guide says Claude won't bypass CAPTCHAs, so you'll be pulled back in to sign in yourself. Never paste a password into the chat.

Mistakes and consistency are the hidden costs

Agents make mistakes. I run them every day and still catch errors in review, like a checkbox clicked twice or a setting dropped into the wrong slot. Those are easy to spot if you know what right looks like. They're invisible if you don't.

Anthropic is also clear that you remain responsible for every action the agent takes in your browser. If it publishes something wrong, that's on your business, not the software.

Then there's consistency, which kills more do-it-yourself marketing than anything else. Week one, you're excited. Week three, a big job lands, two guys call in sick, and the marketing stops. Marketing that runs in bursts doesn't build much. It has to happen every week, including the weeks you're slammed.

I've written about owners who are too busy to market. Agents don't fix that by themselves. They just move the busy part from typing to supervising.

Should you run AI agents yourself? A decision table

Mostly left column? Go for it. Mostly right? Your hour is worth more somewhere else.

If you do it yourself, start with these prompts

If you're going to try it, start with planning, not doing. These two prompts work in Claude or ChatGPT. Fill in the brackets.

If those feel easy and you enjoy the work, keep going. If they feel like a second job, that's useful information too. Here's what to hand off first when you get to that point.