I started a 100% AI business — here’s what happened

There’s a lot of hype around AI and what it can do for business.

Is it hype or reality? How much of AI is actually useful to grow a profitable business?

AI evangelists say AI can do “anything” and “everything”… so, as the GM of a $13 million SaaS business, I wanted to see what “everything” actually means in real business terms:

  1. Could I create a 100% AI business from scratch?
  2. What AI tools and products are good enough to share with my coaching clients?
  3. Is AI all hype, or does it have some actual, practical business value to generate real revenue?

To test the waters, I set out to create a fully-AI business from scratch.

The rules: I wouldn’t use any existing advantages to generate traffic or customers. I’d use publicly-accessible tools. And I’d start a business with an easy entry point that anyone could start.

I wanted to see how much business creation, execution, and growth that my team and I typically do day-to-day could be automated with AI.

At AppSumo Originals, we’ve refined a framework over the years to create million-dollar businesses with a high success rate, limited budget, and a small team:

This strategy hasn’t been a one-hit wonder either. We’ve proven it multiple times with different SaaS products targeting varied markets, demographics, and verticals. It’s worked repeatedly since 2018.

When starting an AI business, I started with the same basic Ideation Framework we use when launching products (like 100,000+ user/$1+ million topline revenue products TidyCal and BreezeDoc):

  1. What market do I know people personally? Having a group of people — friends, former coworkers, alumni, Meetup acquaintances — as initial prospective customers success happen FASTER.
  2. Outside of my first customers, what market has a large-ish potential for growth? I won’t obsess over SAM and TAM and market cap metrics exactly… but I will loosely target something that I think can scale into a million-dollar business.
  3. Do I see easy marketing channels for this product or idea? I want to be able to visualize some growth channels and growth tactics to help with scale and growth once we have product-market fit.

I’m also a big fan of leveraging marketplaces and piggy-backing off of other successful sites.

For example, a Chrome Extension or WordPress plugin gets you access to everyone using Chrome (3+ billion people) or websites built on WordPress (~50% of the Internet). The marketplaces will help you cross-promote your product.

After Ideation Framework, I always recommend executives set S.M.A.R.T. goals for a business. Even for our $13+ million revenue business, our goals, KPIs, and monthly metrics/actions follow a loose S.M.A.R.T. model.

Here’s how I used this same framework for our AI business

STEP #1 – Picking the business idea

To set realistic limits (this was a side hustle I wanted to validate while I ran our Originals business full-time), I picked two key components of the S.M.A.R.T. framework:

  • Achievable: My goal was to break even on my costs in the first month.
  • Time-based: I could dedicate one hour per day to this idea.

I landed on creating a $500 revenue business with 1 hour per day 30 days from launch.

To help me find a S.M.A.R.T. business with these limitations, I prompted ChatGPT:

What stuck out me immediately was an online store. With so many online store platforms, I could create something quick — and leverage an existing marketplace that would help push my product. I decided to go with Etsy because it seemed like the quickest and easiest store to launch.

With the initial “Etsy online store” idea established, I asked ChatGPT my next question:

What is the easiest, most profitable item to sell in an online store?

ChatGPT gave me a few options. I decided on pet supplies, because my girlfriend has a pet pug, I could target her as my first customer, typically people know others like them (i.e. a pug owner knows other pug/dog owners), and they can tell you what websites, forums, etc. they look at (so you can do marketing activities on them).

STEP #2 – Creating the products

With an Etsy store for pug owners picked, I continued prompting ChatGPT to help push my idea along:

  • “What are 10 names for an apparel brand for owners of pugs?” (I picked “WrinkleWear” from the options that ChatGPT provided because it made me laugh and I thought it was catchy.)
  • “Come up with 10 taglines for my apparel site for pug owners and their pugs called ‘WrinkleWear'”
  • Write funny phrases I could put on clothing accessories — like bandanas — for my pug company WrinkleWear”

As I was playing with copy ideas, I started to conceptualize different art and designs. Because of my low-cost requirement, I found an AI art generator on AppSumo called Airbrush.

To help with prompting for quality images, I used a free Stable Diffusion Prompt Generator. To find images I liked I tried a ton of trial-by-error prompts. Sometimes building a business is about brute force.

Some of the example prompts I used that resulted in good images:

cute karate pug with yellow background

cute pug dressed up as a ninja on a brown background, realistic texture, visible stitch line, soft smooth lighting, vibrant studio lighting, modular constructivism, physically based rendering, square image

cute pug in a spacesuit, black background with the milky way, realistic texture, physically based rendering, trending on artstation

Once I had some images that I liked, I started to think about how they would look on products.

I asked ChatGPT to help me create marketing copy on Etsy that could be used for a few products:

  • “Write a description for my t-shirt on Etsy with a cute pug in a spacesuit”
  • “Create a title for my t-shirt on Etsy with a cute pug in a spacesuit”
  • “Write cute personalization instructions (under 256 characters) for a shopper to personalize their pugstronaut t-shirt.”

None of this was meant to be perfect. Instead, it was meant to be 80% “good enough” so I could get products listed quickly — and test for signal/product-market fit (PMF) — before I spent time scaling.

One of the biggest mistakes entrepreneurs make is trying to do everything right away. Get the MVP out there, get users/customers/revenue, iterate… and then scale.

Here’s how the “pugstronaut” design and Etsy listing turned out with AI help:

Using the same methodology (ChatGPT creating copy, Airbrush creating images), I mocked up a handful of product listings across a variety of price points and product categories to see what people liked.

Here’s how the store looked with a few different products:

It’s important to note that none of the products were actually made yet.

I used a t-shirt/goodies creation site to add the designs to products. If anyone ordered my plan was to send them a message that there would be a delay to ship. And in that time, I would order the products from the t-shirt/goodies creation site.

Compare this to how different most entrepreneurs think: They order 200 of each product — costing tons of time and money — before they even make their first sale. Wait for signal before you spend too much time and money.

STEP #3 – The results

AI helped accelerate the launch of a idea/product — by helping write copy, and give me some tips on the market to serve — but it couldn’t actually sell the product.

Reaching out to friends, colleagues, and other prospects still requires manual effort.

And when I reached out to people, I noticed there wasn’t much traction. The market was crowded, people weren’t as interested in spending money as I hoped, and the results weren’t there…

And that’s normal for testing out ideas!

At Sumo Group, we’ve tried tons of products that have failed.

I go into a business idea with a hypothesis, launch a MVP, measure initial traction, and then iterate/scale it if there’s an early signal that shows promise.

This “test before you invest” approach prevents us from spending years of time (and lots of money) developing a product that no one wants.

Whether AI or not, my encouragement when building a business always follows a few general principles:

  1. Launch an idea that solves a need for you, or someone you know and understand well.
  2. Build a MVP/test a simple version of the product to see what they think.
  3. If you see people start to sign up or buy, ask them what they want. It’s important to have a strong line of communication with customers early — even 120,000+ signups later, I still talk to 2-4 TidyCal customers on the phone every month.
  4. Find out where your ideal customers hang out online and offline, and market more to those places. You can just ask your favorite customers what blogs they read, subreddits they check out, and newsletters they like.

Specific to AI, launching a 100% AI-driven business taught me a few things about where we can use AI practically.

Here’s how I’m now using AI in our $13+ million business:

  • I use ChatGPT to help me with data analysis. For example, every week, I dump some CSVs into ChatGPT and ask for data related to KPIs. ChatGPT is really great at data analysis and saving time with crunching numbers. Here’s some example data with KingSumo:
    • I used ChatGPT to help me create templates for our new product BreezeDoc. This saved me time having to manually write common contractor, freelancer, and consultant templates.
    • I’ve used Formula Bot in the past to help me with further data analysis and organization. These days, GPT-4/ChatGPT Data Analyst takes its place, but you may still find Formula Bot useful if you’re doing a lot of number crunching.

    AI can help us in certain areas, but the fundamentals of business are still the same…

    1. Create a product in a market you understand
    2. Find out if people want it — no one cared about business that AI helped me create
    3. If you’re getting signups, interest, or paying customers, focus on actions/improvements for your best customers that help make the product better

    Leave a Reply

    Your email address will not be published. Required fields are marked *