How I Built an AI Assistant for Business to Run Mine in 20 Hours a Week

I run a brand SEO and web design business on 20 hours a week, and I have five employees doing the heavy lifting. Only one of them is human. That’s not a typo. Four of my employees are AI, and they’re the reason I can keep my hours small and my client work strong at the same time.

In this post I’ll show you what an AI assistant for business really looks like when you set it up right, the difference between prompting AI and actually employing it, and the five teammates running my week. Then I’ll show you how to start building your own.

What an AI Assistant for Business Actually Is

Most people think an AI assistant for business means typing a question into a chatbot and copying the answer. That’s part of it, but it’s the smallest part.

A real AI assistant for business has a job. It knows your brand, your rules, your process, and the order you like to do things in. It shows up trained, not blank.

The tool I use for this is Claude, and the trained version of a task lives in something called a skill. Think of a skill as a saved set of instructions your AI runs the same way every single time.

Prompting AI vs Employing AI

Here’s the shift that changed how I work.

When you prompt, you type a request, get an answer, and close the tab. That’s a freelance gig. It’s helpful, but it starts from scratch every time and gives you generic output.

When you employ, each AI worker has a job description, training docs, standing operating procedures, and a list of rules it follows every time. That’s a teammate trained on your business.

A prompt gives you something anyone could get. An employee gives you something only your business would produce. The setup takes time, I won’t pretend it doesn’t. But that upfront work is the whole difference between dabbling with AI and actually scaling with it.

Meet My 4 AI Employees

Employee 1: My Content Director

My content director is a skill trained on my entire podcast and blog publishing process, a 16 page document covering brand voice, SEO rules, internal links, my favorite calls to action, my reading level, and the words I never want used (looking at you, ‘discover’).

It knows the exact order to do everything. It turns my podcast transcript into a blog, writes my Instagram captions, and prompts me with the keyword research I need to go run myself. It does not invent my SEO strategy. I still decide what’s worth writing and which keywords are worth chasing. The strategist’s brain stays in the driver’s seat, and the skill just executes.

This one alone takes my full publishing process for a podcast, blog, YouTube video, and Instagram down to an hour or two.

Employee 2: My Pinterest Manager

My Pinterest manager handles pin batches start to finish. It populates my Notion cards, rotates pin titles across all 52 pages in my Canva template, manages the calls to action, and writes pin descriptions when I want it to. It even studies my blog content and tells me which pins to make next.

It won’t give me a design eye, that part is still mine. But for the heavy, repetitive work of batching a month of pins, it’s a game changer. If Pinterest feels like too much work, this is exactly the kind of system I teach inside my Pinterest marketing course.

Employee 3: My Operations Manager

My operations manager lives in my to-do list. I moved my task management into Notion specifically because it connects to Claude. Now if I think of something while I’m driving, I can pull up a voice note and say, add this to my Notion to-do list, remind me at 6. Done. The thing that used to fall out of my head now lands somewhere I’ll actually see it.

Employee 4: My Personal Assistant

This one is off the clock, but it earns its keep. My personal assistant handles the life admin that drains my energy, mostly meal planning and tracking right now. I talk through my week with it and it builds the plan, and if it doesn’t have the info it needs, it looks it up.

It’s a work in progress, but it’s the only reason I’ve stayed even semi consistent. The point isn’t the meal plans. The point is that the skill of building skills doesn’t stop at the office door.

Employee 5: My One Human Teammate

My fifth employee is the only human, and she’s not really an employee, she’s a contractor. Her name is Nicole, she’s based in the Philippines, and she’s integral to how my business runs.

Nicole works around 10 hours a week, with room to grow. She handles social media, schedules blogs into WordPress once they’re written, edits the podcast, uploads videos to YouTube, and keeps things organized in Notion. She takes almost all of my admin off my plate so I can spend my 20 hours where it counts, on my clients.

AI handles the repeatable tasks. Nicole handles the judgment calls and the hands on work AI can’t touch. That combination is what actually makes the math work.

How I Build a Custom Claude Project for Every Client

Here’s the part I want to be honest about, because I don’t believe in hiding it. I use AI for client work, and my pricing reflects that. Every word on a client’s website is not handwritten by me. I’m not a copywriter, and I’d love for my clients to hire one, but most can’t afford a designer, an SEO strategist, a brand expert, and a copywriter all at once.

So the real choice isn’t AI written or handwritten. It’s AI assisted by someone who knows what they’re doing, or nothing at all. For most small business owners, I’m the reason their vision becomes a real website at a price they can actually pay.

And these aren’t generic skills I copy and paste. Every client gets their own custom Claude project loaded with their brand voice document, their website questionnaire, the SEO research and reporting I personally did, and my copy UX strategy framework for building topical authority. That’s hours of upfront work, and it’s the part most people skip. It’s also why I sell strategy, not just words. You can read more about how I approach that on my SEO services page.

Why Your AI Assistant for Business Still Needs You

You can have the best trained AI employees in the world. If no one tells them what to work on, and you haven’t trained them on your specific zone of genius, it won’t work.

Your job is to decide what gets done. Their job is to do it.

For SEO especially, someone still has to look at the real data, pick the keywords, analyze the competition, and push back when AI suggests a topic that’s already saturated. That part can’t be automated. I always verify my keyword research in Ubersuggest or Google Search Console before I commit, because AI pulls from what already exists, so it can’t hand you a fresh angle. That’s your job. If you want to go deeper on the SEO side, I share a lot of that thinking in my SEO tips.

How to Build a Claude Skill of Your Own

You don’t need to build all five employees at once. Start with one.

Here’s the simple version of how to build a Claude skill that actually helps:

•        Pick one job, not ten. Instagram, or Pinterest, or email. One employee, one focus.

•        Give it training docs. Your brand voice, your past content, your rules, your process.

•        Write down the order you like things done in, and the language you do and don’t use.

•        Save it so you never have to repeat yourself, then call it in when you need that task done.

The narrower the job, the deeper you can train it, and the better the output. That’s the whole secret to AI workflow automation that actually sounds like you.

You don’t need more time, more discipline, or a clone of yourself.

You need a team, and four of those teammates can be AI. Start with one employee. Train it well. Then let it run the task you’re tired of doing yourself.

Want the full story, including the free Pinterest skill I built for my course students? Listen to the full episode below. And if you bought Create It, Pin It, Let It Work, check your email for your new skill.