Why do AI Wrappers get a bad (w)rap?

Why do AI Wrappers get a bad (w)rap?

AI wrappers are tools that send info to an AI API and give you the output. That’s the simple version.

There’s a lot of hate (and jealousy) out there that you shouldn’t build one, they’re not real businesses, you’re not providing value, yadda yadda yadda. 

But, plenty of people are making a ton of money from them. That’s business.

The simplest wrappers provide an interface for you to upload a photo, a PDF, input something and send it to OpenAI (other APIs are available 😉) with instructions (e.g. pull out the data in the pdf into a nice format, what is a formula for finding X in a cell, etc).

Now this kind of business won’t necessarily change the world. But I’d argue that most businesses shouldn’t have that goal anyway. If you want to make a bunch of money, with a small team, from selling a simple thing that you can build quickly, then I’m all for it (it’s what I did). More people should do that instead of raising from VCs, hiring a bunch of people and following the glorified startup dream covered in TechCrunch. 

It could change your world.

If you could make $500k/year from doing this (and you knew how), why wouldn’t you?

Okay, enough talk, let’s look at some examples:

• Formula bot - $200k-300k/year revenue - 1 person team

• PhotoAI - $900k/year revenue - 1 person team

• PDF AI - $500k/year revenue - 1 person team

Let’s break them down;

Formula Bot by David Bressler (link)

• After asking ChatGPT for Excel formula help, he thought a dedicated page could do well for this use case.

• The site started as a simple page with no paywall or log in. 

• He shared it on the Excel subreddit and got interest

• He used no-code (bubble) to go from MVP to real product (with paywall and login)

• He got traffic from Reddit, then grew into influencer marketing (think TikTokers) and an affiliate program

• He only spent $5k on marketing

• Currently makes $300k per year

It’s now a fully-fledged product with all sorts of bells and whistles built around the first very simple ‘AI wrapper’.

The takeaway: You can launch something bare bones, find where your audience is that would find it valuable, iterate on the thing, test payments, and test new features.

Opportunities: You could do the exact same thing for Notion Formulas, Airtable Formulas, and Coda Formulas (seriously, why do they all do their own formulas?!).

Also, ‘Notion templates as a business’ have generated several folks; $300k, $500k, and $1M in ONE year.

Photo AI from Pieter Levels (link)

Firstly, Pieter Levels is a prolific (and successful) founder of many things (as well as countless failed things). But the art is in his launches;

• It started as avatarai. AI tools let you upload photos and get several ai-generated versions, but only if you know how to work with APIs (can code).

• So levels set up a landing page, a typeform to upload your photos and a stripe link to pay. 

• He manually downloaded the photos users uploaded and used the model on replicate to generate new versions for you

• He sent them to his customers

• That validated his idea and has since built a proper product around it, he continuously tunes a model to make images better over time.

• He’s making $80k monthly recurring revenue, with 90%+ profit

• He’s using affiliate programs and influencer marketing to grow but early on he was just tweeting about it.

The takeaway: You only need the most basic MVP to get going - You can build a simple app like this using a form, connect it to Zapier, ping replicate, hugging face, openai etc, and return the output to a customer. Now the world does not need another AI avatar app, but niche it down enough and you end up with something like Headshot Pro from Danny Postma - which is (unsurprisingly) focused on professional headshots. Also making $300k+ a year.

Opportunities: you could make an image restoration site for old photos, an object removal tool, or a product ad builder - many of these have been done, but niche it down to a specific industry that doesn't use AI much and you could find a good opportunity. OpenAI now has a vision API which can ‘understand’ what’s in an image - you could build a simple tool that identifies brands and products in an image. Sounds pointless? Deep.ad built this (pre-OpenAI APIs) and was acquired by private equity last year - likely for single-digit millions or more…

PDF AI by Damon Chen (link)

Damon is a founder and saw this app ‘Looseleaf’ for sale (it was making $0) and was 48 hours old.

• He bought a domain pdf.ai for $10k (not something we can all do, I admit)

• Users upload a pdf and can ask it questions “what does this bit mean” “explain the legal jargon” “what is my termination clause” etc

• He makes $500k/year with over 500k customers

• He uses affiliate programs and influencer marketing

• There are a ton of ai <> pdf options out there, and that’s unsurprising when seeing the search volume

The takeaway: there are plenty of ways to look at what’s out there already and write it off (imagine being the looseleaf founder now!), but figuring out some of the specifics, how it looks and works, how you talk about it etc can make all the difference.

Opportunities: niche down again; bankstatementconverter.com makes $13k/monthly recurring revenue, which does a very similar thing to pdf.ai but it focuses on converting your bank statement (duh!) - detangle.ai focuses on legal documents. Where else are there industries with a lot of jargon or painful ‘reading through docs’ that could be explored? Understanding medical records, insurance claims, terms and conditions, government and policy. The list goes on.

The end.

Let me know if you build one or are thinking of building one now!

Also, I’ve got lots of these kinds of breakdowns and ‘analyses’ (if you can call them that) on my notes app that I’d love to do. Let me know here if you’d be interested in more. (edit: i think the poll below only works if you’re logged in, sorry)