Workflow 30: How to Start a Business With AI: The First Two Weeks

If I had to start a business from scratch, this is the process I'd follow.

I've done it twice. My virtual assistant company came almost naturally, because I was already talking about virtual assistants and it was easy to position myself as someone who hires and manages them.

My AI systems agency was harder. I started it in early 2026, into a market with a lot of agencies already in it. So I had to work out how to stand out, and this is what I learned doing it.

The first two weeks give you three things: proof that people will pay, a clear niche, and the exact words your market uses. Clients come after that, and faster, because you're selling in their language.

What this replaces

Launching with an offer you invented, to a market you guessed at, in words nobody uses.

The instinct when starting something is to be different straight away. Most people try that before they know what already works, and end up explaining a new category to people who didn't ask for one.

Blueprint

Five steps, roughly in order.

Trigger: manual, when you're starting something new. Inputs: your idea, public information about who's already succeeding, and real conversations with your market. Output: a validated offer, a niche, content that ranks for it, and the language your buyers use.

1. Find who's already succeeding at this
2. Research their offers and who they serve, then pick your angle
3. Write the offer and a content plan aimed at your niche
4. Make content, and start showing up where your market is
5. Have real conversations, and collect their words

Organize

What you need:

• Your idea, stated simply
• A sense of which industries you already understand. This is the unfair head start most people ignore
• Somewhere to publish content
• Accounts on the platforms your market actually uses
• Something small you can build and give away as a thank-you

Systematize

1. See what already works

Before anything else, find people already succeeding with a version of your idea. Different niche, different country, different industry, it doesn't matter. You want proof that people pay for this.

That's why virtual assistants worked so well for me for years: people already understood what one was.

I'm thinking of starting a business that [YOUR IDEA].

Find me businesses already succeeding with something similar. They can be
in a different niche, industry, or country. For each:
- What they sell and roughly what it costs
- Who they sell it to
- How long they've been doing it
- Any signs it's working: reviews, repeat customers, visible growth

I'm looking for proof that people pay for this. If you can't find much,
say so plainly, since that's important information.

2. Pick your angle

Then look at their offers closely. You have two ways to stand out: a different offer, or the same offer for a different niche.

For my agency, I saw most people doing systems work for service businesses. I've run an agency myself, so I know their pains. So I picked agencies first. We've shifted since, which is normal. The first niche is where you start learning, not where you end up.

Here are the businesses you found: [PASTE OR REFERENCE].

I have experience in [INDUSTRIES YOU KNOW WELL].

Where's the gap? Either:
- A different offer to the same people, or
- The same offer to a niche they're not serving, ideally one I already
  understand.

Give me two or three options and tell me the tradeoffs of each.

3. Write the offer and the content plan

Have AI draft the offer for your niche, then a list of content to create. Put the niche in the titles.

If you serve med spas, the content says "med spas" in the title. That way, anyone searching for your thing in their industry finds you first.

Write the offer for [SERVICE] for [NICHE].

Then give me 20 content ideas: videos or articles, each titled with
[NICHE] in the name, each answering something that niche actually searches
for.

4. Make the content, and show up

Start publishing. It costs nothing but time, and it becomes a reference you can point to, both for clients and for Claude.

For what it's worth, I still draft my own written pieces, and I hold myself to about a thousand words in half an hour. The videos are me talking. None of them are generated.

At the same time, find your target market where they already are. LinkedIn, Facebook, wherever. Comment on their posts. Ask questions. Be curious, especially if it's an industry you've never been in.

5. Ask to learn, not to sell

Then reach out, and don't pitch. Ask to pick their brain.

Something like: I've been doing this for other industries. I want to understand how it works in yours. Could I ask you a few questions?

A lot of people will say no. Some will say yes. With the ones who do, test the offer: here's what I'm thinking, does this make sense for your industry?

Ask for a referral at the end, and mean it when you say it's fine if they don't have one.

And give them something back. I build small Claude skills or custom GPTs easily, so I'll make one that fits their business as a thank-you. If they actually use it, they remember you. That's what makes referrals happen.

Then, and only then

Once you have a handful of these conversations, you have the exact phrases your market uses to describe the problem. That's when outreach, ads, and everything else start. Now you're paying to reach people with language you know works.

How to set it up

1. Run the "what already works" prompt on your idea. Keep only businesses with real evidence that people pay.
2. Pick your angle: a different offer, or the same offer for a niche you already understand.
3. Draft the offer and twenty content ideas with your niche in each title.
4. Publish consistently, and start commenting where your market already spends time.
5. Ask for short conversations to learn about their industry. Don't pitch. Bring a small thank-you.
6. Write down the exact phrases people use, then start outreach and ads.

Assign

AI does the research, the competitor analysis, the drafting, and the content plan.

The conversations are yours, and they're the part that actually builds the business.

It's worth being clear about why, because the temptation is to skip them. AI can tell you who's succeeding and what they charge. It can draft an offer and twenty titles. What it can't give you is how a person in your target market actually describes their problem at 9pm on a Tuesday, in their own words. That language is the thing that makes everything else work, and you only get it by asking.

The research makes you informed. The conversations make you credible.

Rule of thumb from this build: research what works, then go and ask the people it works for.

Integrate

This workflow borrows pieces from others in the library:

• The competitor research runs the same way as my idea filter
• The content plan uses the same approach as my content planner
• Prospect research before a call uses the same process as Jaq Jaq, my research agent
• The thank-you skill is built the same way as everything else here

Which is the quiet point of this whole series. Starting a business uses the same handful of workflows as running one. You just run them in a different order.

Next: the last one. How to think through building your first full AI automation, start to finish.